Skip to contentSkip to conversation

Mirror allocation policy

linux-lvm Phillip Susi 2 participants 6 messages Jan 5, 2021 – Jan 6, 2021

lore
Share & tools
lore
#0 lore raw

I seem to remember ( and found some references on the web ) that there used to be a way to change the allocation policy when creating a mirror from the default of strict ( must use different pvs ), but I can't find a mention of strict in the man pages these days. I did see the option for --alloc anywhere but lvconvert still refuses to change an lv into a mirror saying that it can't find the extents ( there are plenty of extents, but only one pv ).

How can I forge it to make the mirror on a single pv? Alternatively, how can I create a copy of an lv ( I was planning to mirror, then --splitmirror ).

_______________________________________________
linux-lvm mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/linux-lvm
read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

#1 lore raw
Quoted 1 line · Phillip Susi

How can I forge it to make the mirror on a single pv?

If --alloc anywhere isn't doing the trick, you'll need to dig into the -vvvv output to try to understand why. There might be
a config option setting getting in the way disabling the choice you want it to make, or if it's an algorithmic issue you might try to coax it to do the right thing by removing whatever choice it has in the allocation by specifying the actual pv:extent ranges it must use on the command line.

Alasdair

_______________________________________________
linux-lvm mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/linux-lvm
read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

#2 lore raw
Quoted 7 lines · Alasdair G Kergon
Quoted 1 line · Phillip Susi

How can I forge it to make the mirror on a single pv?

If --alloc anywhere isn't doing the trick, you'll need to dig into the -vvvv output to try to understand why. There might be
a config option setting getting in the way disabling the choice you want it to make, or if it's an algorithmic issue you might try to coax it to

I'm not seeing a whole lot here other than what appears to be this flat out refusal to use extents from the same pv:

#metadata/lv_manip.c:2535 Not using free space on existing
parallel PV /dev/md1.

Limiting the allocation to two specific extent ranges does not help either.

Full log:

#libdm-config.c:1061 Setting
allocation/mirror_logs_require_separate_pvs to 0
#metadata/lv_manip.c:3439 Adjusted allocation request to 7681
logical extents. Existing size 0. New size 7681.
#metadata/lv_manip.c:3442 Mirror log of 1 extents of size 8192
sectors needed for region size 4096.
#libdm-config.c:1061 Setting allocation/maximise_cling to 1
#metadata/pv_map.c:54 Allowing allocation on /dev/md1 start PE
2561 length 23039
#metadata/pv_map.c:54 Allowing allocation on /dev/md1 start PE
128000 length 347469
#metadata/lv_manip.c:2321 Parallel PVs at LE 0 length 7680:
/dev/md1
#metadata/lv_manip.c:3186 Trying allocation using contiguous
policy.
#metadata/lv_manip.c:2788 Areas to be sorted and filled
sequentially.
#metadata/lv_manip.c:2704 Still need 7681 total extents from
370508 remaining (0 positional slots):
#metadata/lv_manip.c:2707 1 (1 data/0 parity) parallel areas
of 7680 extents each
#metadata/lv_manip.c:2711 1 metadata area of 1 extents each
#metadata/lv_manip.c:2535 Not using free space on existing
parallel PV /dev/md1.

 #metadata/lv_manip.c:3186         Trying allocation using cling policy.
 #metadata/lv_manip.c:2783         Cling_to_allocated is set
 #metadata/lv_manip.c:2786         1 preferred area(s) to be filled

positionally.
#metadata/lv_manip.c:2704 Still need 7681 total extents from
370508 remaining (1 positional slots):
#metadata/lv_manip.c:2707 1 (1 data/0 parity) parallel areas
of 7680 extents each
#metadata/lv_manip.c:2711 1 metadata area of 1 extents each
#metadata/lv_manip.c:2535 Not using free space on existing
parallel PV /dev/md1.
#metadata/lv_manip.c:3186 Trying allocation using normal
policy.
#metadata/lv_manip.c:2783 Cling_to_allocated is set
#metadata/lv_manip.c:2788 Areas to be sorted and filled
sequentially.
#metadata/lv_manip.c:2704 Still need 7681 total extents from
370508 remaining (0 positional slots):
#metadata/lv_manip.c:2707 1 (1 data/0 parity) parallel areas
of 7680 extents each
#metadata/lv_manip.c:2711 1 metadata area of 1 extents each
#metadata/lv_manip.c:2535 Not using free space on existing
parallel PV /dev/md1.
#metadata/lv_manip.c:2783 Cling_to_allocated is not set
#metadata/lv_manip.c:2788 Areas to be sorted and filled
sequentially.
#metadata/lv_manip.c:2704 Still need 7681 total extents from
370508 remaining (0 positional slots):
#metadata/lv_manip.c:2707 1 (1 data/0 parity) parallel areas
of 7680 extents each
#metadata/lv_manip.c:2711 1 metadata area of 1 extents each
#metadata/lv_manip.c:2535 Not using free space on existing
parallel PV /dev/md1.
#metadata/lv_manip.c:3220 Insufficient suitable allocatable extents
for logical volume : 7681 more required

_______________________________________________
linux-lvm mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/linux-lvm
read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

#3 lore raw
Quoted 2 lines

#libdm-config.c:1061 Setting
allocation/mirror_logs_require_separate_pvs to 0

So despite that, it's not letting you have the 'anywhere' for the log part of the allocation.

However, for what you are doing, maybe you don't need an on-disk mirror log or can temporarily borrow a little space (add small temporary loopback PV to the VG?) for it?

Quoted 9 lines

#metadata/lv_manip.c:2704 Still need 7681 total extents from
370508 remaining (0 positional slots):
#metadata/lv_manip.c:2707 1 (1 data/0 parity) parallel areas
of 7680 extents each
#metadata/lv_manip.c:2711 1 metadata area of 1 extents each
#metadata/lv_manip.c:2535 Not using free space on existing
parallel PV /dev/md1.
#metadata/lv_manip.c:3220 Insufficient suitable allocatable extents
for logical volume : 7681 more required

Alasdair

_______________________________________________
linux-lvm mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/linux-lvm
read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

#4 lore raw
Quoted 2 lines · Alasdair G Kergon

So despite that, it's not letting you have the 'anywhere' for the log part of the allocation.

Based on the number of extents it said it was looking for, I didn't think it was the log that it couldn't place.

Quoted 3 lines

However, for what you are doing, maybe you don't need an on-disk mirror log or can temporarily borrow a little space (add small temporary loopback PV to the VG?) for it?

I could have sworn that I had tried --corelog yesterday and it still didn't work, but today it did. Weird.

_______________________________________________
linux-lvm mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/linux-lvm
read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

#5 lore raw
Quoted 2 lines · Phillip Susi

Based on the number of extents it said it was looking for, I didn't think it was the log that it couldn't place.

It was looking for what it calls 'parallel' extents i.e. extents not on the same device as the ones belonging to the existing LV being converted, and, at first sight, it wasn't giving you the option of overriding that in respect of the log. Clearly the code could be improved, but it's a very much a niche case, well outside the normal way it's expected to be used.

Alasdair

_______________________________________________
linux-lvm mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/linux-lvm
read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/