Skip to contentSkip to conversation

sco 5(server) linux(client, FC4) cp stalls after a few ten KBs

DISCUSSION Georg Ritter 2 participants 3 messages Dec 13, 2006

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Dear NFS gurus :-),

Problem:
FC4 box as client to a SCO 5 server stops copying after a few 10 KBs, worked on same hardware with RH9 as client, and currently still works for a mdk10 client.

I would be grateful for a hint or better a way to fix it, as I have no idea of what else could be wrong and where/how to investigate it further.

Longer description:
I ran into the following problem, after upgrading from RH9 to FC4 (and also another FC6 box has the same "behaviour") and completely ran out of ideas.

Setup:
FC4/FC6 boxes as clients connecting to an SCO nfs server, small LAN, all class A IPs.

Client can mount the exports from the SCO server, but after a few 10KB of data the speed drops near or to zero and it gets stuck and still seems to copy a few bytes every 30 secs or minute.

Findings so far:
mdk10 laptop as a client can mount the SCO dirs and copies full speed.

It used to work with the exact same hardware but RH9 instead of FC4/FC6.

Looks like a buffer gets filled and then it gets stuck.

nfsd on the SCO don't die away.

No errors on SCO portmap (-v -d) (additionally it worked before, and serves the mdk10 box, without touching the settings on the SCO server)

setting the debugging output on FC4 to 32565 with echo into the according /proc/... vars reveals no obvious error message.

No errors with nfsstat neither on SCO nor FC4.

Network is at good health.

It works the other way round (SCO as client mounting from FC4 box)

Disabled (on FC4) firewall and selinux. No such things on the SCO box.

Seem to be ok in principal because it mounts and copies fine up to the stop after a few ten KBs.

Tried different kernels on the FC4 box (all plain vanilla kernels offered via official update path)

Tried many different options on the mount command: vers=2, nolock different package sizes, forcing tcp etc.: no effect

Thanks,

Georg

P.S. More details from the FC4 (client) box available (logfiles, etc).

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

#1 lore raw
Quoted 8 lines · Georg Ritter

Dear NFS gurus :-),

Problem:
FC4 box as client to a SCO 5 server stops copying after a few 10 KBs, worked on same hardware with RH9 as client, and currently still works for a mdk10 client.

I would be grateful for a hint or better a way to fix it, as I have no idea of what else could be wrong and where/how to investigate it further.

How about starting with a tcpdump of the client and server behaviour? Also, when forcing tcp mounts, you can use 'netstat -t' to monitor the TCP link.

Trond

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

#2 lore raw

VGhhbmtzLCBUcm9uZCwKCmZvciB0aGUgaWRlYS4gSSBkdW1wZWQgdGhlIHRyYWZmaWMsIGRvZXMg aXQgcmV2ZWFsIHF1aWNrbHkgd2hhdCBpcyB3cm9uZywgZm9yPyAKVW5mb3J0dW5hdGxleSwgScK0 bSBub3QgZmlybSBlbm91Z2ggdG8gZGVyaXZlIGFueXRoaW5nIGZyb20gdGhlIGR1bXAuCgpTbyBz ZXR1cCBhZ2FpbjoKbGludXgoRkM0IGJveCwgMi42LjE3LTEuMjE0Ml9GQzRzbXApIGNsaWVudCAo Y2FsbGVkIGdpZ2FieXRlKQogICBtb3VudGluZyBmcm9tIGFuCiJTQ08gT3BlblNlcnZlcihUTSkg UmVsZWFzZSA1IiBhcyBzZXJ2ZXIgKGNhbGxlZCBncCkuCgoidGNwZHVtcCBob3N0IGdwIiAob24g bGludXggY2xpZW50LCBnaWdhYnl0ZSkgb2YgdGhlIGZvbGxvd2luZyBjb21tYW5kcyBpc3N1ZWQ6 Cm1vdW50IC10bmZzIC1vdmVycz0yLHRjcCxub2xvY2sgZ3A6L3Vzci9oYSAvbW50L25mcy9ncC9o YQpjcCAvbW50L25mcy9ncC9oYS8xTUIuZGF0IC90bXAvCmhhbmdzIGZvcmV2ZXIuCmNyZWF0ZXMg YSAiMCAxMy4gRGV6IDE4OjQ1IDFNQi5kYXQiIGluIC90bXAvCgpzb3VyY2UgZmlsZSBAZ3AgaW4g L3Vzci9oYSBpcyAiMTA0ODU3NiBEZXogMTMgMTY6NDAgMU1CLmRhdCIKSXQgbm9ybWFsbHkgYWxz byBjb3BpZXMgc29tZSB0ZW4gS0JzIGFuZCBoYW5ncyBhZnRlciB0aGF0LgoKaGVyZSB0aGUgZW5k IG9mIHRoZSBkdW1wCi4uLgoxOToxMzowNS42MDQwMTggSVAgZ2lnYWJ5dGUuMTAyMyA+IGdwLm5m czogLiBhY2sgMTk5MjUzIHdpbiA2Mjc4MAoxOToxMzowNS42MDQ4OTIgSVAgZ3AubmZzID4gZ2ln YWJ5dGUuMDogcmVwbHkgb2sgOTk2CjE5OjEzOjA1LjYwNDk4NiBJUCBncC5uZnMgPiBnaWdhYnl0 ZS4xMDIzOiAuIGFjayA3MjM3IHdpbiA0MzgwCjE5OjEzOjA1LjYwNjIxNyBJUCBncC5uZnMgPiBn aWdhYnl0ZS4xOTY4NDk1MTU1OiByZXBseSBvayAxNDYwIHJlYWQgW3xuZnNdCjE5OjEzOjA1LjYw NjIyNyBJUCBnaWdhYnl0ZS4xMDIzID4gZ3AubmZzOiAuIGFjayAyMDE3MDkgd2luIDYyNzgwCjE5 OjEzOjA1CgpmdWxsIGR1bXAgaGVyZTogaHR0cDovLzExMjIzMzQ0NTU0NC4yMG0uY29tL25mc2R1 bXAuaHRtbAoKVGh4LAoKCUdlb3JnCgo+IEhvdyBhYm91dCBzdGFydGluZyB3aXRoIGEgdGNwZHVt cCBvZiB0aGUgY2xpZW50IGFuZCBzZXJ2ZXIgYmVoYXZpb3VyPwo+IEFsc28sIHdoZW4gZm9yY2lu ZyB0Y3AgbW91bnRzLCB5b3UgY2FuIHVzZSAnbmV0c3RhdCAtdCcgdG8gbW9uaXRvciB0aGUKPiBU Q1AgbGluay4KPiAKPiBUcm9uZAo+IAo+IAo+IAoKCgotLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tClRha2UgU3Vy dmV5cy4gRWFybiBDYXNoLiBJbmZsdWVuY2UgdGhlIEZ1dHVyZSBvZiBJVApKb2luIFNvdXJjZUZv cmdlLm5ldCdzIFRlY2hzYXkgcGFuZWwgYW5kIHlvdSdsbCBnZXQgdGhlIGNoYW5jZSB0byBzaGFy ZSB5b3VyCm9waW5pb25zIG9uIElUICYgYnVzaW5lc3MgdG9waWNzIHRocm91Z2ggYnJpZWYgc3Vy dmV5cyAtIGFuZCBlYXJuIGNhc2gKaHR0cDovL3d3dy50ZWNoc2F5LmNvbS9kZWZhdWx0LnBocD9w YWdlPWpvaW4ucGhwJnA9c291cmNlZm9yZ2UmQ0lEPURFVkRFVgpfX19fX19fX19fX19fX19fX19f X19fX19fX19fX19fX19fX19fX19fX19fX19fXwpORlMgbWFpbGxpc3QgIC0gIE5GU0BsaXN0cy5z b3VyY2Vmb3JnZS5uZXQKaHR0cHM6Ly9saXN0cy5zb3VyY2Vmb3JnZS5uZXQvbGlzdHMvbGlzdGlu Zm8vbmZzCg==