NFS client request failure statistics?

NFS client request failure statistics?

[Joshua Hoblitt]

[-- Attachment #1.1: Type: text/plain, Size: 782 bytes --]

Hello,

I was wondering if there is some existing means (systemtap, etc.) of
getting a count of NFS client requests that have "failed" after
exceeding the maximum number of retransmissions?  If there is no means
of getting at this information, I feel that it would be worth adding.
It should be a useful stat to have in addition to the retrans count when
tuning the timeo/retrans values.

I've spent a small amount of time poking around in the sources to see
how difficult it would be export this information to userspace.  It
appears that this would be non-trivial to do as it seems that the NFS
client code itself doesn't keep a count of failed requests and the
procfs interface (/proc/net/rpc/nfs) is auto-generated by the sunrpc
layer.  Is my analysis correct?

Cheers,

-J

--

[-- Attachment #1.2: Type: application/pgp-signature, Size: 189 bytes --]

[-- Attachment #2: Type: text/plain, Size: 286 bytes --]

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

[-- Attachment #3: Type: text/plain, Size: 140 bytes --]

_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: NFS client request failure statistics?

[Steve Dickson]

Joshua Hoblitt wrote:
> Hello,
> 
> I was wondering if there is some existing means (systemtap, etc.) of
> getting a count of NFS client requests that have "failed" after
> exceeding the maximum number of retransmissions?  
You can look in systemtap-0.5.13-1.fc6.rpm but I don't think there
are any script that do that...

> If there is no means of getting at this information, I feel that it would be worth adding.
scripts are always welcomed! ;-)

> It should be a useful stat to have in addition to the retrans count when
> tuning the timeo/retrans values.
Also take a look at the NFS/RPC metrics in /proc/self/mounts. They
might have something similar to what your looking for..

> 
> I've spent a small amount of time poking around in the sources to see
> how difficult it would be export this information to userspace.  It
> appears that this would be non-trivial to do as it seems that the NFS
> client code itself doesn't keep a count of failed requests and the
> procfs interface (/proc/net/rpc/nfs) is auto-generated by the sunrpc
> layer.  Is my analysis correct?
Again, I would suggest you look at the /proc/self/mounts interface.

steved.


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: NFS client request failure statistics?

[Chuck Lever]

[-- Attachment #1: Type: text/plain, Size: 1168 bytes --]

Joshua Hoblitt wrote:
> Hello,
> 
> I was wondering if there is some existing means (systemtap, etc.) of
> getting a count of NFS client requests that have "failed" after
> exceeding the maximum number of retransmissions?

Only "soft" mounted NFS file systems will ever "fail" a request after 
some number of retransmits.  "hard" mounted NFS file systems will 
continue retrying forever.  See nfs(5) for more details on using these 
mount options.

The Linux NFS FAQ has some wisdom about why you shouldn't use "soft" 
mounts, but rather "hard,intr".

> I've spent a small amount of time poking around in the sources to see
> how difficult it would be export this information to userspace.  It
> appears that this would be non-trivial to do as it seems that the NFS
> client code itself doesn't keep a count of failed requests and the
> procfs interface (/proc/net/rpc/nfs) is auto-generated by the sunrpc
> layer.  Is my analysis correct?

The traditional statistics-keeping method is in /proc/net/rpc/nfs, but 
there is a new one that is exported in /proc/self/mountstats.  It 
wouldn't be challenging to add a stat there that counted soft requests 
that timed out.

[-- Attachment #2: chuck.lever.vcf --]
[-- Type: text/x-vcard, Size: 291 bytes --]

begin:vcard
fn:Chuck Lever
n:Lever;Chuck
org:Oracle Corporation;Corporate Architecture: Linux Projects Group
adr:;;1015 Granger Avenue;Ann Arbor;MI;48104;USA
title:Principal Member of Staff
tel;work:+1 248 614 5091
x-mozilla-html:FALSE
url:http://oss.oracle.com/~cel/
version:2.1
end:vcard


[-- Attachment #3: Type: text/plain, Size: 286 bytes --]

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

[-- Attachment #4: Type: text/plain, Size: 140 bytes --]

_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs