ceph: flush the delayed caps in time
RFC PATCH 1/1 ceph: flush the delayed caps in time
From: Xiubo Li <[email protected]>
The delayed_work will be executed per 5 seconds, during this time the cap_delay_list may accumulate thounsands of caps need to flush, this will make the MDS's dispatch queue be full and need a very long time to handle them. And if there has some other operations, likes a rmdir request, it will be add in the tail of dispath queue and need to wait for several or tens of seconds.
In client side we shouldn't queue to many of the cap requests and flush them if there has more than 100 items.
URL: https://tracker.ceph.com/issues/51734
--- fs/ceph/caps.c | 21 ++++++++++++++++++++- fs/ceph/mds_client.c | 3 ++- fs/ceph/mds_client.h | 3 +++ 3 files changed, 25 insertions(+), 2 deletions(-)
fs/ceph/caps.c +20 -1
@@ -507,6 +507,8 @@ static void __cap_set_timeouts(struct ceph_mds_client *mdsc,static void __cap_delay_requeue (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num =0 ; +dout ("__cap_delay_requeue %p flags 0x%lx at %lu\n" , &ci->vfs_inode, ci->i_ceph_flags, ci->i_hold_caps_max);if (!mdsc->stopping) {
@@ -515,12 +517,19 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,if (ci->i_ceph_flags &CEPH_I_FLUSH ) goto no_change;list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; }__cap_set_timeouts (mdsc, ci);list_add_tail (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay; no_change:spin_unlock (&mdsc->cap_delay_lock); } + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/*
@@ -531,13 +540,23 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,static void __cap_delay_requeue_front (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num; +dout ("__cap_delay_requeue_front %p\n" , &ci->vfs_inode);spin_lock (&mdsc->cap_delay_lock); ci->i_ceph_flags |=CEPH_I_FLUSH ; -if (!list_empty (&ci->i_cap_delay_list)) +if (!list_empty (&ci->i_cap_delay_list)) {list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; + }list_add (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay;spin_unlock (&mdsc->cap_delay_lock); + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/*
fs/ceph/mds_client.c +2 -1
@@ -4514,7 +4514,7 @@ void inc_session_sequence(struct ceph_mds_session *s)/* * delayed work -- periodically trim expired leases, renew caps with mds */ -static void schedule_delayed (struct ceph_mds_client *mdsc) +void schedule_delayed (struct ceph_mds_client *mdsc) {int delay =5 ;unsigned hz =round_jiffies_relative (HZ * delay);
@@ -4616,6 +4616,7 @@ int ceph_mdsc_init(struct ceph_fs_client *fsc) mdsc->request_tree =RB_ROOT ;INIT_DELAYED_WORK (&mdsc->delayed_work, delayed_work); mdsc->last_renew_caps = jiffies; + mdsc->num_cap_delay =0 ;INIT_LIST_HEAD (&mdsc->cap_delay_list);INIT_LIST_HEAD (&mdsc->cap_wait_list);spin_lock_init (&mdsc->cap_delay_lock);
fs/ceph/mds_client.h +3 -0
@@ -423,6 +423,7 @@ struct ceph_mds_client {struct rb_root request_tree;/* pending mds requests */ struct delayed_work delayed_work;/* delayed work */ unsigned long last_renew_caps;/* last time we renewed our caps */ +unsigned long num_cap_delay;/* caps in the cap_delay_list */ struct list_head cap_delay_list;/* caps with delayed release */ spinlock_t cap_delay_lock;/* protects cap_delay_list */ struct list_head snap_flush_list;/* cap_snaps ready to flush */
@@ -568,6 +569,8 @@ extern int ceph_trim_caps(struct ceph_mds_client *mdsc,struct ceph_mds_session *session, int max_caps); +extern void schedule_delayed (struct ceph_mds_client *mdsc); +static inline int ceph_wait_on_async_create (struct inode *inode) {struct ceph_inode_info *ci =ceph_inode (inode);
Signature · 1 line
2.27.0
Quoted 9 lines · On Wed, 2021-07-21 at 16:27 +0800, [email protected] wrote:
From: Xiubo Li <[email protected]>
The
delayed_workwill be executed per 5 seconds, during this time thecap_delay_listmay accumulate thounsands of caps need to flush, this will make the MDS's dispatch queue be full and need a very long time to handle them. And if there has some other operations, likes a rmdir request, it will be add in the tail of dispath queue and need to wait for several or tens of seconds.
Quoted 3 lines
In client side we shouldn't queue to many of the cap requests and flush them if there has more than 100 items.
Why 100? My inclination is to say NAK on this.
I'm not a fan of this sort of arbitrary throttling in order to alleviate a scalability problem in the MDS. The cap delay logic is already horrifically complex as well, and this just makes it worse. If the MDS happens to be processing requests from a lot of clients, this may not help at all.
Quoted 1 line
The core problem sounds like the MDS got a call and just took way too long to get to it. Shouldn't we be addressing this in the MDS instead? It sounds like it needs to do a better job of parallelizing cap flushes.
Quoted 107 lines
Signed-off-by Xiubo Li <[email protected]>--- fs/ceph/caps.c | 21 ++++++++++++++++++++- fs/ceph/mds_client.c | 3 ++- fs/ceph/mds_client.h | 3 +++ 3 files changed, 25 insertions(+), 2 deletions(-)fs/ceph/caps.c +20
-1@@ -507,6 +507,8 @@ static void __cap_set_timeouts(struct ceph_mds_client *mdsc,static void __cap_delay_requeue (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num =0 ; +dout ("__cap_delay_requeue %p flags 0x%lx at %lu\n" , &ci->vfs_inode, ci->i_ceph_flags, ci->i_hold_caps_max);if (!mdsc->stopping) {@@ -515,12 +517,19 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,if (ci->i_ceph_flags &CEPH_I_FLUSH ) goto no_change;list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; }__cap_set_timeouts (mdsc, ci);list_add_tail (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay; no_change:spin_unlock (&mdsc->cap_delay_lock); } + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* @@ -531,13 +540,23 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,static void __cap_delay_requeue_front (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num; +dout ("__cap_delay_requeue_front %p\n" , &ci->vfs_inode);spin_lock (&mdsc->cap_delay_lock); ci->i_ceph_flags |=CEPH_I_FLUSH ; -if (!list_empty (&ci->i_cap_delay_list)) +if (!list_empty (&ci->i_cap_delay_list)) {list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; + }list_add (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay;spin_unlock (&mdsc->cap_delay_lock); + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* fs/ceph/mds_client.c +2
-1@@ -4514,7 +4514,7 @@ void inc_session_sequence(struct ceph_mds_session *s)/* * delayed work -- periodically trim expired leases, renew caps with mds */ -static void schedule_delayed (struct ceph_mds_client *mdsc) +void schedule_delayed (struct ceph_mds_client *mdsc) {int delay =5 ;unsigned hz =round_jiffies_relative (HZ * delay);@@ -4616,6 +4616,7 @@ int ceph_mdsc_init(struct ceph_fs_client *fsc) mdsc->request_tree =RB_ROOT ;INIT_DELAYED_WORK (&mdsc->delayed_work, delayed_work); mdsc->last_renew_caps = jiffies; + mdsc->num_cap_delay =0 ;INIT_LIST_HEAD (&mdsc->cap_delay_list);INIT_LIST_HEAD (&mdsc->cap_wait_list);spin_lock_init (&mdsc->cap_delay_lock);fs/ceph/mds_client.h +3
-0@@ -423,6 +423,7 @@ struct ceph_mds_client {struct rb_root request_tree;/* pending mds requests */ struct delayed_work delayed_work;/* delayed work */ unsigned long last_renew_caps;/* last time we renewed our caps */ +unsigned long num_cap_delay;/* caps in the cap_delay_list */ struct list_head cap_delay_list;/* caps with delayed release */ spinlock_t cap_delay_lock;/* protects cap_delay_list */ struct list_head snap_flush_list;/* cap_snaps ready to flush */ @@ -568,6 +569,8 @@ extern int ceph_trim_caps(struct ceph_mds_client *mdsc,struct ceph_mds_session *session, int max_caps); +extern void schedule_delayed (struct ceph_mds_client *mdsc); +static inline int ceph_wait_on_async_create (struct inode *inode) {struct ceph_inode_info *ci =ceph_inode (inode);
Signature · 1 line
Jeff Layton <[email protected]>
Quoted 10 lines · Jeff Layton
Quoted 9 lines · On Wed, 2021-07-21 at 16:27 +0800, [email protected] wrote:
From: Xiubo Li <[email protected]>
The
delayed_workwill be executed per 5 seconds, during this time thecap_delay_listmay accumulate thounsands of caps need to flush, this will make the MDS's dispatch queue be full and need a very long time to handle them. And if there has some other operations, likes a rmdir request, it will be add in the tail of dispath queue and need to wait for several or tens of seconds.
Quoted 4 lines
Quoted 3 lines
In client side we shouldn't queue to many of the cap requests and flush them if there has more than 100 items.
Why 100? My inclination is to say NAK on this.
This just from my test that around 100 client_caps requests queued will work fine in most cases, which won't take too long to handle. Some times the client will send thousands of requests in a short time, that will be a problem.
Quoted 6 lines
I'm not a fan of this sort of arbitrary throttling in order to alleviate a scalability problem in the MDS. The cap delay logic is already horrifically complex as well, and this just makes it worse. If the MDS happens to be processing requests from a lot of clients, this may not help at all.
Yeah, right, in that case the mds dispatch queue possibly will have more in a short time.
Quoted 4 lines
Quoted 1 line
The core problem sounds like the MDS got a call and just took way too long to get to it. Shouldn't we be addressing this in the MDS instead? It sounds like it needs to do a better job of parallelizing cap flushes.
The problem is that almost all the requests need to acquire the global 'mds_lock', so the parallelizing won't improve much.
Currently all the client side requests will have the priority of CEPH_MSG_PRIO_DEFAULT:
m->hdr.priority = cpu_to_le16(CEPH_MSG_PRIO_DEFAULT);
Not sure whether can we just make the cap flushing request a lower priority ? If this won't introduce other potential issue, that will work because in the MDS dispatch queue it will handle high priority requests first.
Thanks
Quoted 107 lines
Quoted 107 lines
Signed-off-by Xiubo Li <[email protected]>--- fs/ceph/caps.c | 21 ++++++++++++++++++++- fs/ceph/mds_client.c | 3 ++- fs/ceph/mds_client.h | 3 +++ 3 files changed, 25 insertions(+), 2 deletions(-)fs/ceph/caps.c +20
-1@@ -507,6 +507,8 @@ static void __cap_set_timeouts(struct ceph_mds_client *mdsc,static void __cap_delay_requeue (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num =0 ; +dout ("__cap_delay_requeue %p flags 0x%lx at %lu\n" , &ci->vfs_inode, ci->i_ceph_flags, ci->i_hold_caps_max);if (!mdsc->stopping) {@@ -515,12 +517,19 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,if (ci->i_ceph_flags &CEPH_I_FLUSH ) goto no_change;list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; }__cap_set_timeouts (mdsc, ci);list_add_tail (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay; no_change:spin_unlock (&mdsc->cap_delay_lock); } + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* @@ -531,13 +540,23 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,static void __cap_delay_requeue_front (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num; +dout ("__cap_delay_requeue_front %p\n" , &ci->vfs_inode);spin_lock (&mdsc->cap_delay_lock); ci->i_ceph_flags |=CEPH_I_FLUSH ; -if (!list_empty (&ci->i_cap_delay_list)) +if (!list_empty (&ci->i_cap_delay_list)) {list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; + }list_add (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay;spin_unlock (&mdsc->cap_delay_lock); + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* fs/ceph/mds_client.c +2
-1@@ -4514,7 +4514,7 @@ void inc_session_sequence(struct ceph_mds_session *s)/* * delayed work -- periodically trim expired leases, renew caps with mds */ -static void schedule_delayed (struct ceph_mds_client *mdsc) +void schedule_delayed (struct ceph_mds_client *mdsc) {int delay =5 ;unsigned hz =round_jiffies_relative (HZ * delay);@@ -4616,6 +4616,7 @@ int ceph_mdsc_init(struct ceph_fs_client *fsc) mdsc->request_tree =RB_ROOT ;INIT_DELAYED_WORK (&mdsc->delayed_work, delayed_work); mdsc->last_renew_caps = jiffies; + mdsc->num_cap_delay =0 ;INIT_LIST_HEAD (&mdsc->cap_delay_list);INIT_LIST_HEAD (&mdsc->cap_wait_list);spin_lock_init (&mdsc->cap_delay_lock);fs/ceph/mds_client.h +3
-0@@ -423,6 +423,7 @@ struct ceph_mds_client {struct rb_root request_tree;/* pending mds requests */ struct delayed_work delayed_work;/* delayed work */ unsigned long last_renew_caps;/* last time we renewed our caps */ +unsigned long num_cap_delay;/* caps in the cap_delay_list */ struct list_head cap_delay_list;/* caps with delayed release */ spinlock_t cap_delay_lock;/* protects cap_delay_list */ struct list_head snap_flush_list;/* cap_snaps ready to flush */ @@ -568,6 +569,8 @@ extern int ceph_trim_caps(struct ceph_mds_client *mdsc,struct ceph_mds_session *session, int max_caps); +extern void schedule_delayed (struct ceph_mds_client *mdsc); +static inline int ceph_wait_on_async_create (struct inode *inode) {struct ceph_inode_info *ci =ceph_inode (inode);
Quoted 14 lines · Xiubo Li
Quoted 13 lines · Jeff Layton
Quoted 12 lines · On Wed, 2021-07-21 at 16:27 +0800, [email protected] wrote:
From: Xiubo Li <[email protected]>
The
delayed_workwill be executed per 5 seconds, during this time thecap_delay_listmay accumulate thounsands of caps need to flush, this will make the MDS's dispatch queue be full and need a very long time to handle them. And if there has some other operations, likes a rmdir request, it will be add in the tail of dispath queue and need to wait for several or tens of seconds.In client side we shouldn't queue to many of the cap requests and flush them if there has more than 100 items.
Quoted 6 lines
Quoted 1 line
Why 100? My inclination is to say NAK on this.
This just from my test that around 100
client_capsrequests queued will work fine in most cases, which won't take too long to handle. Some times the client will send thousands of requests in a short time, that will be a problem.
What may be a better approach is to figure out why we're holding on to so many caps and trying to flush them all at once. Maybe if we were to more aggressively flush sooner, we'd not end up with such a backlog?
Quoted 10 lines
Quoted 5 lines
I'm not a fan of this sort of arbitrary throttling in order to alleviate a scalability problem in the MDS. The cap delay logic is already horrifically complex as well, and this just makes it worse. If the MDS happens to be processing requests from a lot of clients, this may not help at all.
Yeah, right, in that case the mds dispatch queue possibly will have more in a short time.
Quoted 8 lines
Quoted 4 lines
Quoted 1 line
The core problem sounds like the MDS got a call and just took way too long to get to it. Shouldn't we be addressing this in the MDS instead? It sounds like it needs to do a better job of parallelizing cap flushes.
The problem is that almost all the requests need to acquire the global '
mds_lock', so the parallelizing won't improve much.
Yuck. That sounds like the root cause right there. For the record, I'm against papering over that with client patches.
Quoted 10 lines
Currently all the client side requests will have the priority of
CEPH_MSG_PRIO_DEFAULT:
m->hdr.priority=cpu_to_le16(CEPH_MSG_PRIO_DEFAULT);Not sure whether can we just make the cap flushing request a lower priority ? If this won't introduce other potential issue, that will work because in the MDS dispatch queue it will handle high priority requests first.
Tough call. You might mark it for lower priority, but then someone calls fsync() and suddenly you need it to finish sooner. I think that sort of change might backfire in some cases.
Usually when faced with this sort of issue, the right answer is to try to work ahead of the "spikes" in activity so that you don't have quite such a backlog when you do need to flush everything out. I'm not sure how possible that is here.
One underlying issue may be that some of the client's background activity is currently driven by arbitrary timers (e.g. the delayed_work timer). When the timer pops, it suddenly has to do a bunch of work, when some of it could have been done earlier so that the load was spread out.
Quoted 108 lines
Quoted 107 lines
Quoted 107 lines
Signed-off-by Xiubo Li <[email protected]>--- fs/ceph/caps.c | 21 ++++++++++++++++++++- fs/ceph/mds_client.c | 3 ++- fs/ceph/mds_client.h | 3 +++ 3 files changed, 25 insertions(+), 2 deletions(-)fs/ceph/caps.c +20
-1@@ -507,6 +507,8 @@ static void __cap_set_timeouts(struct ceph_mds_client *mdsc,static void __cap_delay_requeue (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num =0 ; +dout ("__cap_delay_requeue %p flags 0x%lx at %lu\n" , &ci->vfs_inode, ci->i_ceph_flags, ci->i_hold_caps_max);if (!mdsc->stopping) {@@ -515,12 +517,19 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,if (ci->i_ceph_flags &CEPH_I_FLUSH ) goto no_change;list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; }__cap_set_timeouts (mdsc, ci);list_add_tail (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay; no_change:spin_unlock (&mdsc->cap_delay_lock); } + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* @@ -531,13 +540,23 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,static void __cap_delay_requeue_front (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num; +dout ("__cap_delay_requeue_front %p\n" , &ci->vfs_inode);spin_lock (&mdsc->cap_delay_lock); ci->i_ceph_flags |=CEPH_I_FLUSH ; -if (!list_empty (&ci->i_cap_delay_list)) +if (!list_empty (&ci->i_cap_delay_list)) {list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; + }list_add (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay;spin_unlock (&mdsc->cap_delay_lock); + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* fs/ceph/mds_client.c +2
-1@@ -4514,7 +4514,7 @@ void inc_session_sequence(struct ceph_mds_session *s)/* * delayed work -- periodically trim expired leases, renew caps with mds */ -static void schedule_delayed (struct ceph_mds_client *mdsc) +void schedule_delayed (struct ceph_mds_client *mdsc) {int delay =5 ;unsigned hz =round_jiffies_relative (HZ * delay);@@ -4616,6 +4616,7 @@ int ceph_mdsc_init(struct ceph_fs_client *fsc) mdsc->request_tree =RB_ROOT ;INIT_DELAYED_WORK (&mdsc->delayed_work, delayed_work); mdsc->last_renew_caps = jiffies; + mdsc->num_cap_delay =0 ;INIT_LIST_HEAD (&mdsc->cap_delay_list);INIT_LIST_HEAD (&mdsc->cap_wait_list);spin_lock_init (&mdsc->cap_delay_lock);fs/ceph/mds_client.h +3
-0@@ -423,6 +423,7 @@ struct ceph_mds_client {struct rb_root request_tree;/* pending mds requests */ struct delayed_work delayed_work;/* delayed work */ unsigned long last_renew_caps;/* last time we renewed our caps */ +unsigned long num_cap_delay;/* caps in the cap_delay_list */ struct list_head cap_delay_list;/* caps with delayed release */ spinlock_t cap_delay_lock;/* protects cap_delay_list */ struct list_head snap_flush_list;/* cap_snaps ready to flush */ @@ -568,6 +569,8 @@ extern int ceph_trim_caps(struct ceph_mds_client *mdsc,struct ceph_mds_session *session, int max_caps); +extern void schedule_delayed (struct ceph_mds_client *mdsc); +static inline int ceph_wait_on_async_create (struct inode *inode) {struct ceph_inode_info *ci =ceph_inode (inode);
Signature · 1 line
Jeff Layton <[email protected]>
Quoted 15 lines · Jeff Layton
Quoted 14 lines · Xiubo Li
Quoted 13 lines · Jeff Layton
Quoted 12 lines · On Wed, 2021-07-21 at 16:27 +0800, [email protected] wrote:
From: Xiubo Li <[email protected]>
The
delayed_workwill be executed per 5 seconds, during this time thecap_delay_listmay accumulate thounsands of caps need to flush, this will make the MDS's dispatch queue be full and need a very long time to handle them. And if there has some other operations, likes a rmdir request, it will be add in the tail of dispath queue and need to wait for several or tens of seconds.In client side we shouldn't queue to many of the cap requests and flush them if there has more than 100 items.
Quoted 8 lines
Quoted 5 lines
Quoted 1 line
Why 100? My inclination is to say NAK on this.
This just from my test that around 100
client_capsrequests queued will work fine in most cases, which won't take too long to handle. Some times the client will send thousands of requests in a short time, that will be a problem.What may be a better approach is to figure out why we're holding on to so many caps and trying to flush them all at once. Maybe if we were to more aggressively flush sooner, we'd not end up with such a backlog?
Yeah, I am still working on this.
From my analysis, some fixing or improvement like this will help reduce it:
commit e64f44a884657358812e6c057957be546db03cbe
Author: Xiubo Li <[email protected]>
Date: Wed May 27 09:09:27 2020 -0400
ceph: skip checking caps when session reconnecting and releasing reqs
It make no sense to check the caps when reconnecting to mds. And for the async dirop caps, they will be put by its _cb() function, so when releasing the requests, it will make no sense too.
From the test result, it seems that some caps related requests were still pending in the cap_delay_list even after the rmdir requests were already sent out and finished for the previous directories. I am still confirming this.
Quoted 25 lines
Quoted 10 lines
Quoted 6 lines
I'm not a fan of this sort of arbitrary throttling in order to alleviate a scalability problem in the MDS. The cap delay logic is already horrifically complex as well, and this just makes it worse. If the MDS happens to be processing requests from a lot of clients, this may not help at all.
Yeah, right, in that case the mds dispatch queue possibly will have more in a short time.
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
The core problem sounds like the MDS got a call and just took way too long to get to it. Shouldn't we be addressing this in the MDS instead? It sounds like it needs to do a better job of parallelizing cap flushes.
The problem is that almost all the requests need to acquire the global '
mds_lock', so the parallelizing won't improve much.Yuck. That sounds like the root cause right there. For the record, I'm against papering over that with client patches.
Quoted 5 lines
Currently all the client side requests will have the priority of
CEPH_MSG_PRIO_DEFAULT:
m->hdr.priority=cpu_to_le16(CEPH_MSG_PRIO_DEFAULT);
Quoted 8 lines
Quoted 5 lines
Not sure whether can we just make the cap flushing request a lower priority ? If this won't introduce other potential issue, that will work because in the MDS dispatch queue it will handle high priority requests first.
Tough call. You might mark it for lower priority, but then someone calls
fsync()and suddenly you need it to finish sooner. I think that sort of change might backfire in some cases.
Yeah, I am also worrying about this.
Quoted 119 lines
Usually when faced with this sort of issue, the right answer is to try to work ahead of the "spikes" in activity so that you don't have quite such a backlog when you do need to flush everything out. I'm not sure how possible that is here.
One underlying issue may be that some of the client's background activity is currently driven by arbitrary timers (e.g. the
delayed_worktimer). When the timer pops, it suddenly has to do a bunch of work, when some of it could have been done earlier so that the load was spread out.Quoted 107 lines
Quoted 107 lines
Quoted 107 lines
Signed-off-by Xiubo Li <[email protected]>--- fs/ceph/caps.c | 21 ++++++++++++++++++++- fs/ceph/mds_client.c | 3 ++- fs/ceph/mds_client.h | 3 +++ 3 files changed, 25 insertions(+), 2 deletions(-)fs/ceph/caps.c +20
-1@@ -507,6 +507,8 @@ static void __cap_set_timeouts(struct ceph_mds_client *mdsc,static void __cap_delay_requeue (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num =0 ; +dout ("__cap_delay_requeue %p flags 0x%lx at %lu\n" , &ci->vfs_inode, ci->i_ceph_flags, ci->i_hold_caps_max);if (!mdsc->stopping) {@@ -515,12 +517,19 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,if (ci->i_ceph_flags &CEPH_I_FLUSH ) goto no_change;list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; }__cap_set_timeouts (mdsc, ci);list_add_tail (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay; no_change:spin_unlock (&mdsc->cap_delay_lock); } + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* @@ -531,13 +540,23 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,static void __cap_delay_requeue_front (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num; +dout ("__cap_delay_requeue_front %p\n" , &ci->vfs_inode);spin_lock (&mdsc->cap_delay_lock); ci->i_ceph_flags |=CEPH_I_FLUSH ; -if (!list_empty (&ci->i_cap_delay_list)) +if (!list_empty (&ci->i_cap_delay_list)) {list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; + }list_add (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay;spin_unlock (&mdsc->cap_delay_lock); + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* fs/ceph/mds_client.c +2
-1@@ -4514,7 +4514,7 @@ void inc_session_sequence(struct ceph_mds_session *s)/* * delayed work -- periodically trim expired leases, renew caps with mds */ -static void schedule_delayed (struct ceph_mds_client *mdsc) +void schedule_delayed (struct ceph_mds_client *mdsc) {int delay =5 ;unsigned hz =round_jiffies_relative (HZ * delay);@@ -4616,6 +4616,7 @@ int ceph_mdsc_init(struct ceph_fs_client *fsc) mdsc->request_tree =RB_ROOT ;INIT_DELAYED_WORK (&mdsc->delayed_work, delayed_work); mdsc->last_renew_caps = jiffies; + mdsc->num_cap_delay =0 ;INIT_LIST_HEAD (&mdsc->cap_delay_list);INIT_LIST_HEAD (&mdsc->cap_wait_list);spin_lock_init (&mdsc->cap_delay_lock);fs/ceph/mds_client.h +3
-0@@ -423,6 +423,7 @@ struct ceph_mds_client {struct rb_root request_tree;/* pending mds requests */ struct delayed_work delayed_work;/* delayed work */ unsigned long last_renew_caps;/* last time we renewed our caps */ +unsigned long num_cap_delay;/* caps in the cap_delay_list */ struct list_head cap_delay_list;/* caps with delayed release */ spinlock_t cap_delay_lock;/* protects cap_delay_list */ struct list_head snap_flush_list;/* cap_snaps ready to flush */ @@ -568,6 +569,8 @@ extern int ceph_trim_caps(struct ceph_mds_client *mdsc,struct ceph_mds_session *session, int max_caps); +extern void schedule_delayed (struct ceph_mds_client *mdsc); +static inline int ceph_wait_on_async_create (struct inode *inode) {struct ceph_inode_info *ci =ceph_inode (inode);
Quoted 35 lines · Xiubo Li
Quoted 15 lines · Jeff Layton
Quoted 14 lines · Xiubo Li
Quoted 13 lines · Jeff Layton
Quoted 12 lines · On Wed, 2021-07-21 at 16:27 +0800, [email protected] wrote:
From: Xiubo Li <[email protected]>
The
delayed_workwill be executed per 5 seconds, during this time thecap_delay_listmay accumulate thounsands of caps need to flush, this will make the MDS's dispatch queue be full and need a very long time to handle them. And if there has some other operations, likes a rmdir request, it will be add in the tail of dispath queue and need to wait for several or tens of seconds.In client side we shouldn't queue to many of the cap requests and flush them if there has more than 100 items.
Quoted 8 lines
Quoted 5 lines
Quoted 1 line
Why 100? My inclination is to say NAK on this.
This just from my test that around 100
client_capsrequests queued will work fine in most cases, which won't take too long to handle. Some times the client will send thousands of requests in a short time, that will be a problem.What may be a better approach is to figure out why we're holding on to so many caps and trying to flush them all at once. Maybe if we were to more aggressively flush sooner, we'd not end up with such a backlog?
Yeah, I am still working on this.
From my analysis, some fixing or improvement like this will help reduce it:
commit
e64f44a884657358812e6c057957be546db03cbe
Author: Xiubo Li <[email protected]>
Date: Wed May 27 09:09:27 2020 -0400ceph: skip checking caps when session reconnecting and releasing reqs
Quoted 9 lines
It make no sense to check the caps when reconnecting to mds. And for the async dirop caps, they will be put by its
_cb()function, so when releasing the requests, it will make no sense too.From the test result, it seems that some caps related requests were
still pending in thecap_delay_listeven after the rmdir requests were already sent out and finished for the previous directories. I am still confirming this.
Ok, that at least sounds superficially similar to this bug that Patrick and I have been chasing:
https://tracker.ceph.com/issues/51279
It's possible that some of these have the same root cause.
Quoted 159 lines
Quoted 27 lines
Quoted 12 lines
Quoted 8 lines
I'm not a fan of this sort of arbitrary throttling in order to alleviate a scalability problem in the MDS. The cap delay logic is already horrifically complex as well, and this just makes it worse. If the MDS happens to be processing requests from a lot of clients, this may not help at all.
Yeah, right, in that case the mds dispatch queue possibly will have more in a short time.
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
The core problem sounds like the MDS got a call and just took way too long to get to it. Shouldn't we be addressing this in the MDS instead? It sounds like it needs to do a better job of parallelizing cap flushes.
The problem is that almost all the requests need to acquire the global '
mds_lock', so the parallelizing won't improve much.Yuck. That sounds like the root cause right there. For the record, I'm against papering over that with client patches.
Quoted 5 lines
Currently all the client side requests will have the priority of
CEPH_MSG_PRIO_DEFAULT:
m->hdr.priority=cpu_to_le16(CEPH_MSG_PRIO_DEFAULT);Quoted 8 lines
Quoted 5 lines
Not sure whether can we just make the cap flushing request a lower priority ? If this won't introduce other potential issue, that will work because in the MDS dispatch queue it will handle high priority requests first.
Tough call. You might mark it for lower priority, but then someone calls
fsync()and suddenly you need it to finish sooner. I think that sort of change might backfire in some cases.Yeah, I am also worrying about this.
Quoted 119 lines
Usually when faced with this sort of issue, the right answer is to try to work ahead of the "spikes" in activity so that you don't have quite such a backlog when you do need to flush everything out. I'm not sure how possible that is here.
One underlying issue may be that some of the client's background activity is currently driven by arbitrary timers (e.g. the
delayed_worktimer). When the timer pops, it suddenly has to do a bunch of work, when some of it could have been done earlier so that the load was spread out.Quoted 107 lines
Quoted 107 lines
Quoted 107 lines
Signed-off-by Xiubo Li <[email protected]>--- fs/ceph/caps.c | 21 ++++++++++++++++++++- fs/ceph/mds_client.c | 3 ++- fs/ceph/mds_client.h | 3 +++ 3 files changed, 25 insertions(+), 2 deletions(-)fs/ceph/caps.c +20
-1@@ -507,6 +507,8 @@ static void __cap_set_timeouts(struct ceph_mds_client *mdsc,static void __cap_delay_requeue (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num =0 ; +dout ("__cap_delay_requeue %p flags 0x%lx at %lu\n" , &ci->vfs_inode, ci->i_ceph_flags, ci->i_hold_caps_max);if (!mdsc->stopping) {@@ -515,12 +517,19 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,if (ci->i_ceph_flags &CEPH_I_FLUSH ) goto no_change;list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; }__cap_set_timeouts (mdsc, ci);list_add_tail (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay; no_change:spin_unlock (&mdsc->cap_delay_lock); } + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* @@ -531,13 +540,23 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,static void __cap_delay_requeue_front (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num; +dout ("__cap_delay_requeue_front %p\n" , &ci->vfs_inode);spin_lock (&mdsc->cap_delay_lock); ci->i_ceph_flags |=CEPH_I_FLUSH ; -if (!list_empty (&ci->i_cap_delay_list)) +if (!list_empty (&ci->i_cap_delay_list)) {list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; + }list_add (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay;spin_unlock (&mdsc->cap_delay_lock); + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* fs/ceph/mds_client.c +2
-1@@ -4514,7 +4514,7 @@ void inc_session_sequence(struct ceph_mds_session *s)/* * delayed work -- periodically trim expired leases, renew caps with mds */ -static void schedule_delayed (struct ceph_mds_client *mdsc) +void schedule_delayed (struct ceph_mds_client *mdsc) {int delay =5 ;unsigned hz =round_jiffies_relative (HZ * delay);@@ -4616,6 +4616,7 @@ int ceph_mdsc_init(struct ceph_fs_client *fsc) mdsc->request_tree =RB_ROOT ;INIT_DELAYED_WORK (&mdsc->delayed_work, delayed_work); mdsc->last_renew_caps = jiffies; + mdsc->num_cap_delay =0 ;INIT_LIST_HEAD (&mdsc->cap_delay_list);INIT_LIST_HEAD (&mdsc->cap_wait_list);spin_lock_init (&mdsc->cap_delay_lock);fs/ceph/mds_client.h +3
-0@@ -423,6 +423,7 @@ struct ceph_mds_client {struct rb_root request_tree;/* pending mds requests */ struct delayed_work delayed_work;/* delayed work */ unsigned long last_renew_caps;/* last time we renewed our caps */ +unsigned long num_cap_delay;/* caps in the cap_delay_list */ struct list_head cap_delay_list;/* caps with delayed release */ spinlock_t cap_delay_lock;/* protects cap_delay_list */ struct list_head snap_flush_list;/* cap_snaps ready to flush */ @@ -568,6 +569,8 @@ extern int ceph_trim_caps(struct ceph_mds_client *mdsc,struct ceph_mds_session *session, int max_caps); +extern void schedule_delayed (struct ceph_mds_client *mdsc); +static inline int ceph_wait_on_async_create (struct inode *inode) {struct ceph_inode_info *ci =ceph_inode (inode);
Signature · 1 line
Jeff Layton <[email protected]>
Quoted 15 lines · Jeff Layton
Quoted 14 lines · Xiubo Li
Quoted 13 lines · Jeff Layton
Quoted 12 lines · On Wed, 2021-07-21 at 16:27 +0800, [email protected] wrote:
From: Xiubo Li <[email protected]>
The
delayed_workwill be executed per 5 seconds, during this time thecap_delay_listmay accumulate thounsands of caps need to flush, this will make the MDS's dispatch queue be full and need a very long time to handle them. And if there has some other operations, likes a rmdir request, it will be add in the tail of dispath queue and need to wait for several or tens of seconds.In client side we shouldn't queue to many of the cap requests and flush them if there has more than 100 items.
Quoted 8 lines
Quoted 5 lines
Quoted 1 line
Why 100? My inclination is to say NAK on this.
This just from my test that around 100
client_capsrequests queued will work fine in most cases, which won't take too long to handle. Some times the client will send thousands of requests in a short time, that will be a problem.What may be a better approach is to figure out why we're holding on to so many caps and trying to flush them all at once. Maybe if we were to more aggressively flush sooner, we'd not end up with such a backlog?
881912 Jul 22 10:36:14 lxbceph1 kernel: ceph: 00000000f7ee4ccf mode 040755 uid.gid 0.0
881913 Jul 22 10:36:14 lxbceph1 kernel: ceph: size 0 -> 0
881914 Jul 22 10:36:14 lxbceph1 kernel: ceph: truncate_seq 0 -> 1
881915 Jul 22 10:36:14 lxbceph1 kernel: ceph: truncate_size 0 -> 18446744073709551615
881916 Jul 22 10:36:14 lxbceph1 kernel: ceph: add_cap 00000000f7ee4ccf mds0 cap 152d pAsxLsXsxFsx seq 1
881917 Jul 22 10:36:14 lxbceph1 kernel: ceph: lookup_snap_realm 1 0000000095ff27ff
881918 Jul 22 10:36:14 lxbceph1 kernel: ceph: get_realm 0000000095ff27ff 5421 -> 5422
881919 Jul 22 10:36:14 lxbceph1 kernel: ceph: __ceph_caps_issued 00000000f7ee4ccf cap 000000003c8bc134 issued -
881920 Jul 22 10:36:14 lxbceph1 kernel: ceph: marking 00000000f7ee4ccf NOT complete
881921 Jul 22 10:36:14 lxbceph1 kernel: ceph: issued pAsxLsXsxFsx, mds wanted -, actual -, queueing
881922 Jul 22 10:36:14 lxbceph1 kernel: ceph: __cap_set_timeouts 00000000f7ee4ccf min 5036 max 60036
881923 Jul 22 10:36:14 lxbceph1 kernel: ceph: __cap_delay_requeue 00000000f7ee4ccf flags 0 at 4294896928
881924 Jul 22 10:36:14 lxbceph1 kernel: ceph: add_cap inode 00000000f7ee4ccf (1000000152a.fffffffffffffffe) cap 000000003c8bc134 pAsxLsXsxFsx now pAsxLsXsxFsx seq 1 mds0 881925 Jul 22 10:36:14 lxbceph1 kernel: ceph: marking 00000000f7ee4ccf complete (empty)
881926 Jul 22 10:36:14 lxbceph1 kernel: ceph: dn 0000000079fd7e04 attached to 00000000f7ee4ccf ino 1000000152a.fffffffffffffffe
881927 Jul 22 10:36:14 lxbceph1 kernel: ceph: update_dentry_lease 0000000079fd7e04 duration 30000 ms ttl 4294866888
881928 Jul 22 10:36:14 lxbceph1 kernel: ceph: dentry_lru_touch 000000006fddf4b0 0000000079fd7e04 'removalC.test01806' (offset 0)
From the kernel logs, some of these delayed caps are from previous thousands of 'mkdir' requests, after the mkdir requests get replied and when creating new caps, since the MDS has issued extra caps that they don't want, then these caps are added to the delay queue, and in 5 seconds the delay list could accumulate up to several thousands of caps.
And most of them come from stale dentries.
So 5 seconds later when the delayed_queue is fired, all this client_caps requests are queued in the MDS dispatch queue.
The following commit could improve the performance very much:
commit 37c4efc1ddf98ba8b234d116d863a9464445901e
Author: Yan, Zheng <[email protected]>
Date: Thu Jan 31 16:55:51 2019 +0800
ceph: periodically trim stale dentries
Previous commit make VFS delete stale dentry when last reference is dropped. Lease also can become invalid when corresponding dentry has no reference. This patch make cephfs periodically scan lease list, delete corresponding dentry if lease is invalid.
There are two types of lease, dentry lease and dir lease. dentry lease has life time and applies to singe dentry. Dentry lease is added to tail of a list when it's updated, leases at front of the list will expire first. Dir lease is CEPH_CAP_FILE_SHARED on directory inode, it applies to all dentries in the directory. Dentries have dir leases are added to another list. Dentries in the list are periodically checked in a round robin manner.
With this commit, it will take 3~4 minutes to finish my test:
real 3m33.998s
user 0m0.644s
sys 0m2.341s
[1] Done ( cd /mnt/kcephfs.$i && time strace -Tv -o ~/removal${i}.log -- rm -rf removal$i* ) real 3m34.028s
user 0m0.620s
sys 0m2.342s
[2]- Done ( cd /mnt/kcephfs.$i && time strace -Tv -o ~/removal${i}.log -- rm -rf removal$i* ) real 3m34.049s
user 0m0.638s
sys 0m2.315s
[3]+ Done ( cd /mnt/kcephfs.$i && time strace -Tv -o ~/removal${i}.log -- rm -rf removal$i* )
Without this it will take more than 12 minutes for above 3 'rm' threads.
Thanks.
Quoted 151 lines
Quoted 9 lines
Quoted 5 lines
I'm not a fan of this sort of arbitrary throttling in order to alleviate a scalability problem in the MDS. The cap delay logic is already horrifically complex as well, and this just makes it worse. If the MDS happens to be processing requests from a lot of clients, this may not help at all.
Yeah, right, in that case the mds dispatch queue possibly will have more in a short time.
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
The core problem sounds like the MDS got a call and just took way too long to get to it. Shouldn't we be addressing this in the MDS instead? It sounds like it needs to do a better job of parallelizing cap flushes.
The problem is that almost all the requests need to acquire the global '
mds_lock', so the parallelizing won't improve much.Yuck. That sounds like the root cause right there. For the record, I'm against papering over that with client patches.
Quoted 10 lines
Currently all the client side requests will have the priority of
CEPH_MSG_PRIO_DEFAULT:
m->hdr.priority=cpu_to_le16(CEPH_MSG_PRIO_DEFAULT);Not sure whether can we just make the cap flushing request a lower priority ? If this won't introduce other potential issue, that will work because in the MDS dispatch queue it will handle high priority requests first.
Tough call. You might mark it for lower priority, but then someone calls
fsync()and suddenly you need it to finish sooner. I think that sort of change might backfire in some cases.Usually when faced with this sort of issue, the right answer is to try to work ahead of the "spikes" in activity so that you don't have quite such a backlog when you do need to flush everything out. I'm not sure how possible that is here.
One underlying issue may be that some of the client's background activity is currently driven by arbitrary timers (e.g. the
delayed_worktimer). When the timer pops, it suddenly has to do a bunch of work, when some of it could have been done earlier so that the load was spread out.Quoted 107 lines
Quoted 107 lines
Quoted 107 lines
Signed-off-by Xiubo Li <[email protected]>--- fs/ceph/caps.c | 21 ++++++++++++++++++++- fs/ceph/mds_client.c | 3 ++- fs/ceph/mds_client.h | 3 +++ 3 files changed, 25 insertions(+), 2 deletions(-)fs/ceph/caps.c +20
-1@@ -507,6 +507,8 @@ static void __cap_set_timeouts(struct ceph_mds_client *mdsc,static void __cap_delay_requeue (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num =0 ; +dout ("__cap_delay_requeue %p flags 0x%lx at %lu\n" , &ci->vfs_inode, ci->i_ceph_flags, ci->i_hold_caps_max);if (!mdsc->stopping) {@@ -515,12 +517,19 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,if (ci->i_ceph_flags &CEPH_I_FLUSH ) goto no_change;list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; }__cap_set_timeouts (mdsc, ci);list_add_tail (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay; no_change:spin_unlock (&mdsc->cap_delay_lock); } + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* @@ -531,13 +540,23 @@ static void __cap_delay_requeue(struct ceph_mds_client *mdsc,static void __cap_delay_requeue_front (struct ceph_mds_client *mdsc,struct ceph_inode_info *ci) { +int num; +dout ("__cap_delay_requeue_front %p\n" , &ci->vfs_inode);spin_lock (&mdsc->cap_delay_lock); ci->i_ceph_flags |=CEPH_I_FLUSH ; -if (!list_empty (&ci->i_cap_delay_list)) +if (!list_empty (&ci->i_cap_delay_list)) {list_del_init (&ci->i_cap_delay_list); + mdsc->num_cap_delay--; + }list_add (&ci->i_cap_delay_list, &mdsc->cap_delay_list); + num = ++mdsc->num_cap_delay;spin_unlock (&mdsc->cap_delay_lock); + +if (num >100 ) { +flush_delayed_work (&mdsc->delayed_work); +schedule_delayed (mdsc); + } }/* fs/ceph/mds_client.c +2
-1@@ -4514,7 +4514,7 @@ void inc_session_sequence(struct ceph_mds_session *s)/* * delayed work -- periodically trim expired leases, renew caps with mds */ -static void schedule_delayed (struct ceph_mds_client *mdsc) +void schedule_delayed (struct ceph_mds_client *mdsc) {int delay =5 ;unsigned hz =round_jiffies_relative (HZ * delay);@@ -4616,6 +4616,7 @@ int ceph_mdsc_init(struct ceph_fs_client *fsc) mdsc->request_tree =RB_ROOT ;INIT_DELAYED_WORK (&mdsc->delayed_work, delayed_work); mdsc->last_renew_caps = jiffies; + mdsc->num_cap_delay =0 ;INIT_LIST_HEAD (&mdsc->cap_delay_list);INIT_LIST_HEAD (&mdsc->cap_wait_list);spin_lock_init (&mdsc->cap_delay_lock);fs/ceph/mds_client.h +3
-0@@ -423,6 +423,7 @@ struct ceph_mds_client {struct rb_root request_tree;/* pending mds requests */ struct delayed_work delayed_work;/* delayed work */ unsigned long last_renew_caps;/* last time we renewed our caps */ +unsigned long num_cap_delay;/* caps in the cap_delay_list */ struct list_head cap_delay_list;/* caps with delayed release */ spinlock_t cap_delay_lock;/* protects cap_delay_list */ struct list_head snap_flush_list;/* cap_snaps ready to flush */ @@ -568,6 +569,8 @@ extern int ceph_trim_caps(struct ceph_mds_client *mdsc,struct ceph_mds_session *session, int max_caps); +extern void schedule_delayed (struct ceph_mds_client *mdsc); +static inline int ceph_wait_on_async_create (struct inode *inode) {struct ceph_inode_info *ci =ceph_inode (inode);
Quoted 111 lines · Xiubo Li
Quoted 15 lines · Jeff Layton
Quoted 14 lines · Xiubo Li
Quoted 13 lines · Jeff Layton
Quoted 12 lines · On Wed, 2021-07-21 at 16:27 +0800, [email protected] wrote:
From: Xiubo Li <[email protected]>
The
delayed_workwill be executed per 5 seconds, during this time thecap_delay_listmay accumulate thounsands of caps need to flush, this will make the MDS's dispatch queue be full and need a very long time to handle them. And if there has some other operations, likes a rmdir request, it will be add in the tail of dispath queue and need to wait for several or tens of seconds.In client side we shouldn't queue to many of the cap requests and flush them if there has more than 100 items.
Quoted 8 lines
Quoted 5 lines
Quoted 1 line
Why 100? My inclination is to say NAK on this.
This just from my test that around 100
client_capsrequests queued will work fine in most cases, which won't take too long to handle. Some times the client will send thousands of requests in a short time, that will be a problem.What may be a better approach is to figure out why we're holding on to so many caps and trying to flush them all at once. Maybe if we were to more aggressively flush sooner, we'd not end up with such a backlog?
881912 Jul 22 10:36:14 lxbceph1 kernel: ceph:
00000000f7ee4ccfmode
040755 uid.gid 0.0
881913 Jul 22 10:36:14 lxbceph1 kernel: ceph: size 0 -> 0 881914 Jul 22 10:36:14 lxbceph1 kernel: ceph:truncate_seq0 -> 1 881915 Jul 22 10:36:14 lxbceph1 kernel: ceph:truncate_size0 ->
18446744073709551615
881916 Jul 22 10:36:14 lxbceph1 kernel: ceph:add_cap00000000f7ee4ccfmds0 cap 152d pAsxLsXsxFsx seq 1
881917 Jul 22 10:36:14 lxbceph1 kernel: ceph:lookup_snap_realm10000000095ff27ff
881918 Jul 22 10:36:14 lxbceph1 kernel: ceph:get_realm0000000095ff27ff5421 -> 5422
881919 Jul 22 10:36:14 lxbceph1 kernel: ceph:__ceph_caps_issued00000000f7ee4ccfcap000000003c8bc134issued -
881920 Jul 22 10:36:14 lxbceph1 kernel: ceph: marking00000000f7ee4ccfNOT complete
881921 Jul 22 10:36:14 lxbceph1 kernel: ceph: issued pAsxLsXsxFsx,
mds wanted -, actual -, queueing
881922 Jul 22 10:36:14 lxbceph1 kernel: ceph:__cap_set_timeouts00000000f7ee4ccfmin 5036 max 60036
881923 Jul 22 10:36:14 lxbceph1 kernel: ceph:__cap_delay_requeue00000000f7ee4ccfflags 0 at 4294896928
881924 Jul 22 10:36:14 lxbceph1 kernel: ceph:add_capinode00000000f7ee4ccf(1000000152a.fffffffffffffffe) cap000000003c8bc134pAsxLsXsxFsx now pAsxLsXsxFsx seq 1 mds0
881925 Jul 22 10:36:14 lxbceph1 kernel: ceph: marking00000000f7ee4ccfcomplete (empty)
881926 Jul 22 10:36:14 lxbceph1 kernel: ceph: dn0000000079fd7e04
attached to00000000f7ee4ccfino 1000000152a.fffffffffffffffe
881927 Jul 22 10:36:14 lxbceph1 kernel: ceph:update_dentry_lease0000000079fd7e04duration 30000 ms ttl 4294866888
881928 Jul 22 10:36:14 lxbceph1 kernel: ceph:dentry_lru_touch000000006fddf4b00000000079fd7e04'removalC.test01806' (offset 0)From the kernel logs, some of these delayed caps are from previous
thousands of 'mkdir' requests, after the mkdir requests get replied and when creating new caps, since the MDS has issued extra caps that they don't want, then these caps are added to the delay queue, and in 5 seconds the delay list could accumulate up to several thousands of caps.And most of them come from stale dentries.
So 5 seconds later when the
delayed_queueis fired, all thisclient_capsrequests are queued in the MDS dispatch queue.The following commit could improve the performance very much:
commit
37c4efc1ddf98ba8b234d116d863a9464445901e
Author: Yan, Zheng <[email protected]>
Date: Thu Jan 31 16:55:51 2019 +0800ceph: periodically trim stale dentries
Previous commit make VFS delete stale dentry when last reference is dropped. Lease also can become invalid when corresponding dentry has no reference. This patch make cephfs periodically scan lease list, delete corresponding dentry if lease is invalid.
There are two types of lease, dentry lease and dir lease. dentry lease has life time and applies to singe dentry. Dentry lease is added to
tail
of a list when it's updated, leases at front of the list will expire first. Dir lease isCEPH_CAP_FILE_SHAREDon directory inode, it applies to all dentries in the directory. Dentries have dir leases are added to another list. Dentries in the list are periodically checked in a round robin manner.With this commit, it will take 3~4 minutes to finish my test:
real 3m33.998s
user 0m0.644s
sys 0m2.341s
[1] Done ( cd/mnt/kcephfs.$i&& time strace -Tv -o ~/removal${i}.log -- rm -rf removal$i* )
real 3m34.028s
user 0m0.620s
sys 0m2.342s
[2]- Done ( cd/mnt/kcephfs.$i&& time strace -Tv -o ~/removal${i}.log -- rm -rf removal$i* )
real 3m34.049s
user 0m0.638s
sys 0m2.315s
[3]+ Done ( cd/mnt/kcephfs.$i&& time strace -Tv -o ~/removal${i}.log -- rm -rf removal$i* )
Quoted 2 lines
Without this it will take more than 12 minutes for above 3 'rm' threads.
That commit was already merged a couple of years ago. Does this mean you think the client behavior can't be improved further?
Signature · 1 line
Jeff Layton <[email protected]>
Quoted 111 lines · Jeff Layton
Quoted 110 lines · Xiubo Li
Quoted 15 lines · Jeff Layton
Quoted 14 lines · Xiubo Li
Quoted 13 lines · Jeff Layton
Quoted 12 lines · On Wed, 2021-07-21 at 16:27 +0800, [email protected] wrote:
From: Xiubo Li <[email protected]>
The
delayed_workwill be executed per 5 seconds, during this time thecap_delay_listmay accumulate thounsands of caps need to flush, this will make the MDS's dispatch queue be full and need a very long time to handle them. And if there has some other operations, likes a rmdir request, it will be add in the tail of dispath queue and need to wait for several or tens of seconds.In client side we shouldn't queue to many of the cap requests and flush them if there has more than 100 items.
Quoted 8 lines
Quoted 5 lines
Quoted 1 line
Why 100? My inclination is to say NAK on this.
This just from my test that around 100
client_capsrequests queued will work fine in most cases, which won't take too long to handle. Some times the client will send thousands of requests in a short time, that will be a problem.What may be a better approach is to figure out why we're holding on to so many caps and trying to flush them all at once. Maybe if we were to more aggressively flush sooner, we'd not end up with such a backlog?
881912 Jul 22 10:36:14 lxbceph1 kernel: ceph:
00000000f7ee4ccfmode
040755 uid.gid 0.0
881913 Jul 22 10:36:14 lxbceph1 kernel: ceph: size 0 -> 0 881914 Jul 22 10:36:14 lxbceph1 kernel: ceph:truncate_seq0 -> 1 881915 Jul 22 10:36:14 lxbceph1 kernel: ceph:truncate_size0 ->
18446744073709551615
881916 Jul 22 10:36:14 lxbceph1 kernel: ceph:add_cap00000000f7ee4ccfmds0 cap 152d pAsxLsXsxFsx seq 1
881917 Jul 22 10:36:14 lxbceph1 kernel: ceph:lookup_snap_realm10000000095ff27ff
881918 Jul 22 10:36:14 lxbceph1 kernel: ceph:get_realm0000000095ff27ff5421 -> 5422
881919 Jul 22 10:36:14 lxbceph1 kernel: ceph:__ceph_caps_issued00000000f7ee4ccfcap000000003c8bc134issued -
881920 Jul 22 10:36:14 lxbceph1 kernel: ceph: marking00000000f7ee4ccfNOT complete
881921 Jul 22 10:36:14 lxbceph1 kernel: ceph: issued pAsxLsXsxFsx,
mds wanted -, actual -, queueing
881922 Jul 22 10:36:14 lxbceph1 kernel: ceph:__cap_set_timeouts00000000f7ee4ccfmin 5036 max 60036
881923 Jul 22 10:36:14 lxbceph1 kernel: ceph:__cap_delay_requeue00000000f7ee4ccfflags 0 at 4294896928
881924 Jul 22 10:36:14 lxbceph1 kernel: ceph:add_capinode00000000f7ee4ccf(1000000152a.fffffffffffffffe) cap000000003c8bc134pAsxLsXsxFsx now pAsxLsXsxFsx seq 1 mds0
881925 Jul 22 10:36:14 lxbceph1 kernel: ceph: marking00000000f7ee4ccfcomplete (empty)
881926 Jul 22 10:36:14 lxbceph1 kernel: ceph: dn0000000079fd7e04
attached to00000000f7ee4ccfino 1000000152a.fffffffffffffffe
881927 Jul 22 10:36:14 lxbceph1 kernel: ceph:update_dentry_lease0000000079fd7e04duration 30000 ms ttl 4294866888
881928 Jul 22 10:36:14 lxbceph1 kernel: ceph:dentry_lru_touch000000006fddf4b00000000079fd7e04'removalC.test01806' (offset 0)From the kernel logs, some of these delayed caps are from previous
thousands of 'mkdir' requests, after the mkdir requests get replied and when creating new caps, since the MDS has issued extra caps that they don't want, then these caps are added to the delay queue, and in 5 seconds the delay list could accumulate up to several thousands of caps.And most of them come from stale dentries.
So 5 seconds later when the
delayed_queueis fired, all thisclient_capsrequests are queued in the MDS dispatch queue.The following commit could improve the performance very much:
commit
37c4efc1ddf98ba8b234d116d863a9464445901e
Author: Yan, Zheng <[email protected]>
Date: Thu Jan 31 16:55:51 2019 +0800ceph: periodically trim stale dentries
Previous commit make VFS delete stale dentry when last reference is dropped. Lease also can become invalid when corresponding dentry has no reference. This patch make cephfs periodically scan lease list, delete corresponding dentry if lease is invalid.
There are two types of lease, dentry lease and dir lease. dentry lease has life time and applies to singe dentry. Dentry lease is added to
tail
of a list when it's updated, leases at front of the list will expire first. Dir lease isCEPH_CAP_FILE_SHAREDon directory inode, it applies to all dentries in the directory. Dentries have dir leases are added to another list. Dentries in the list are periodically checked in a round robin manner.With this commit, it will take 3~4 minutes to finish my test:
real 3m33.998s
user 0m0.644s
sys 0m2.341s
[1] Done ( cd/mnt/kcephfs.$i&& time strace -Tv -o ~/removal${i}.log -- rm -rf removal$i* )
real 3m34.028s
user 0m0.620s
sys 0m2.342s
[2]- Done ( cd/mnt/kcephfs.$i&& time strace -Tv -o ~/removal${i}.log -- rm -rf removal$i* )
real 3m34.049s
user 0m0.638s
sys 0m2.315s
[3]+ Done ( cd/mnt/kcephfs.$i&& time strace -Tv -o ~/removal${i}.log -- rm -rf removal$i* )
Quoted 4 lines
Quoted 2 lines
Without this it will take more than 12 minutes for above 3 'rm' threads.
That commit was already merged a couple of years ago. Does this mean you think the client behavior can't be improved further?
No, I think we can.
The customer is mainly using the downstream kernel which doesn't include it yet.
I have patched it and with this it can save a lot of time for my test. In theory and in rare case from my test that the delayed_work() still could send a large number of cap requests in a short time. It's very hard for me to reproduce with this fixing together with the Finisher's fix in https://github.com/ceph/ceph/pull/42394.
I checked that downstream code, there has some buggy codes which could impact the perf, but in the following numerous downstream patches they have been removed or fixed by coincidence when fixing other issues. In the upstream kernel I didn't find any new issue yet.
Thanks.