RE: interprtation of /proc/net/rpc/nfsd

RE: interprtation of /proc/net/rpc/nfsd

[Lever, Charles]

> Well... zeroing out the nfs stats was something I proposed a=20
> while ago but got shoot down both internally and externally=20
> since zero out stats is "not the Linux way".... so I've taken=20
> the patch out of both FC and RHEL kernels, but I guess I=20
> forgot to take it out of the nfs-utils....

for the record, the reason that zeroing the kernel stats is a bad idea
is that this should really be a user-space thing.  otherwise, if process
A is sampling the stats and process B zeroes them, it will screw up
process A's samples.

we can get all the functionality we need by teaching the user-space
tools to subtract.  :-)


-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: interprtation of /proc/net/rpc/nfsd

[Steve Dickson]

Lever, Charles wrote:

>>Well... zeroing out the nfs stats was something I proposed a 
>>while ago but got shoot down both internally and externally 
>>since zero out stats is "not the Linux way".... so I've taken 
>>the patch out of both FC and RHEL kernels, but I guess I 
>>forgot to take it out of the nfs-utils....
>>    
>>
>
>for the record, the reason that zeroing the kernel stats is a bad idea
>is that this should really be a user-space thing.  otherwise, if process
>A is sampling the stats and process B zeroes them, it will screw up
>process A's samples.
>  
>
This make senses for stats that measure time based things like bit rate or
latency.... but for stats that measure simple activity, sampling does not
really come into play.... imho...

SteveD.


-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

interprtation of /proc/net/rpc/nfsd

[Jehan PROCACCIA]

Hello,

I am tunning my NFS configuration on a RedHat ES v3 (equivalent Fedora 
Core 1 , kernel 2.4.X)  and by reading the docs (NFS howto perf 5.6 )  I 
should see if I have enough NFSD daemon by looking at  
/proc/net/rpc/nfs, especialy line "th"

$ grep th /proc/net/rpc/nfsd
th 8 41598 20357.020 3974.540 1809.300 0.000 998.210 781.650 483.840 
116.640 0.000 825.230

However I cannot understand how to read that information !?
I see that I have 8 threads (8 nfsd daemon runing) but figures after ???

Second question,

How on an NFS client (Here I have fedora core 2 clients) can I see how 
is currently mounted an NFS filsystem, I want to check what are the 
current options in use (nfs version, rsize, wsize, sync/async, hard, 
soft etc ...). On the server I can have a look at /var/lib/nfs/xtab

$ cat /var/lib/nfs/xtab | grep arvouin
/mci/mci        
arvouin.int-evry.fr(rw,async,wdelay,hide,nocrossmnt,secure,root_squash,no_all_squash,subtree_check,secure_locks,acl,mapping=identity,anonuid=-2,anongid=-2)

but I am often changing options on the server (for tunning purpose) , 
and hence I want to check that my clients is mounted with the latest 
options,  do I need to remount on the client everytime I change options 
on the server ? (exportfs -r -v -a)  ... .checking what is the current 
options on the client would be more clear to me .

thanks.

PS other lines are also not clear in /proc/net/rpc/nfsd
are they clearly explain somewhere ? or is there a better tool to read 
them ?

$ cat /proc/net/rpc/nfsd
rc 77 2279806 4791576
fh 36 7103818 0 107 461
io 2157379507 3834347693
th 8 41598 20368.720 3975.220 1809.300 0.000 998.210 781.650 483.840 
116.640 0.000 825.230
ra 16 316291 228 80 73 25 40 42 17 36 7 22131
net 7071459 56823 7014636 155
rpc 7071459 0 0 0 0
proc2 18 95 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
proc3 22 107 2606933 544561 626398 870018 178 338970 1189408 253948 
14640 200 0 219018 12808 45181 119 0 66482 4981 22 0 277392




-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: interprtation of /proc/net/rpc/nfsd

[James Pearson]

Jehan PROCACCIA wrote:
> Hello,
> 
> I am tunning my NFS configuration on a RedHat ES v3 (equivalent Fedora 
> Core 1 , kernel 2.4.X)  and by reading the docs (NFS howto perf 5.6 )  I 
> should see if I have enough NFSD daemon by looking at  
> /proc/net/rpc/nfs, especialy line "th"
> 
> $ grep th /proc/net/rpc/nfsd
> th 8 41598 20357.020 3974.540 1809.300 0.000 998.210 781.650 483.840 
> 116.640 0.000 825.230
> 
> However I cannot understand how to read that information !?
> I see that I have 8 threads (8 nfsd daemon runing) but figures after ???

See:

http://nfs.sourceforge.net/nfs-howto/performance.html#NFSD-INSTANCE

> Second question,
> 
> How on an NFS client (Here I have fedora core 2 clients) can I see how 
> is currently mounted an NFS filsystem, I want to check what are the 
> current options in use (nfs version, rsize, wsize, sync/async, hard, 
> soft etc ...). On the server I can have a look at /var/lib/nfs/xtab
> 
> $ cat /var/lib/nfs/xtab | grep arvouin
> /mci/mci        
> arvouin.int-evry.fr(rw,async,wdelay,hide,nocrossmnt,secure,root_squash,no_all_squash,subtree_check,secure_locks,acl,mapping=identity,anonuid=-2,anongid=-2) 
> 
> 
> but I am often changing options on the server (for tunning purpose) , 
> and hence I want to check that my clients is mounted with the latest 
> options,  do I need to remount on the client everytime I change options 
> on the server ? (exportfs -r -v -a)  ... .checking what is the current 
> options on the client would be more clear to me .

I think you have to remount.

Have a look at /proc/mounts on the client.

James Pearson


-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: interprtation of /proc/net/rpc/nfsd

[Jehan PROCACCIA]

James Pearson wrote:

> Jehan PROCACCIA wrote:
>
>> Hello,
>>
>> I am tunning my NFS configuration on a RedHat ES v3 (equivalent 
>> Fedora Core 1 , kernel 2.4.X)  and by reading the docs (NFS howto 
>> perf 5.6 )  I should see if I have enough NFSD daemon by looking at  
>> /proc/net/rpc/nfs, especialy line "th"
>>
>> $ grep th /proc/net/rpc/nfsd
>> th 8 41598 20357.020 3974.540 1809.300 0.000 998.210 781.650 483.840 
>> 116.640 0.000 825.230
>>
>> However I cannot understand how to read that information !?
>> I see that I have 8 threads (8 nfsd daemon runing) but figures after ???
>
>
> See:
>
> http://nfs.sourceforge.net/nfs-howto/performance.html#NFSD-INSTANCE

Yes, I've read that part , but I cannot understand it :-(
The last ten numbers on the *th* line in that file indicate the number 
of seconds that the thread usage was at that percentage of the maximum 
allowable.

$ grep th /proc/net/rpc/nfsd
th 8 41598 20357.020 3974.540 1809.300 0.000 998.210 781.650 483.840 
116.640 0.000 825.230

Why ten numbers, I have 8 thraeds ! ? here for the last number for 
exemple, I had 825seconds that thread number 8 was at 230% of it's 
maximum ???

>
>> Second question,
>>
>> How on an NFS client (Here I have fedora core 2 clients) can I see 
>> how is currently mounted an NFS filsystem, I want to check what are 
>> the current options in use (nfs version, rsize, wsize, sync/async, 
>> hard, soft etc ...). On the server I can have a look at 
>> /var/lib/nfs/xtab
>>
>> $ cat /var/lib/nfs/xtab | grep arvouin
>> /mci/mci        
>> arvouin.int-evry.fr(rw,async,wdelay,hide,nocrossmnt,secure,root_squash,no_all_squash,subtree_check,secure_locks,acl,mapping=identity,anonuid=-2,anongid=-2) 
>>
>>
>> but I am often changing options on the server (for tunning purpose) , 
>> and hence I want to check that my clients is mounted with the latest 
>> options,  do I need to remount on the client everytime I change 
>> options on the server ? (exportfs -r -v -a)  ... .checking what is 
>> the current options on the client would be more clear to me .
>
>
> I think you have to remount.
>
> Have a look at /proc/mounts on the client.
>
Great , that nearly what I wanted
[root@arvouin nfs]# grep procacci /proc/mounts
cobra3:/mci/mci/procacci /mci/mci/procacci nfs 
rw,v3,rsize=8192,wsize=8192,soft,intr,tcp,lock,addr=cobra3 0 0

can't see however if export is using sync or async :-( .
thanks a lot.

> James Pearson




-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: interprtation of /proc/net/rpc/nfsd

[James Pearson]

Jehan PROCACCIA wrote:
> Yes, I've read that part , but I cannot understand it :-(
> The last ten numbers on the *th* line in that file indicate the number 
> of seconds that the thread usage was at that percentage of the maximum 
> allowable.
> 
> $ grep th /proc/net/rpc/nfsd
> th 8 41598 20357.020 3974.540 1809.300 0.000 998.210 781.650 483.840 
> 116.640 0.000 825.230
> 
> Why ten numbers, I have 8 thraeds ! ? here for the last number for 
> exemple, I had 825seconds that thread number 8 was at 230% of it's 
> maximum ???

The following post from Neil Brown has a 'better' explanation:

http://marc.theaimsgroup.com/?l=linux-nfs&m=102824853219024&w=2

James Pearson


-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: interprtation of /proc/net/rpc/nfsd

[Jehan PROCACCIA]

James Pearson wrote:

>>
>> $ grep th /proc/net/rpc/nfsd
>> th 8 41598 20357.020 3974.540 1809.300 0.000 998.210 781.650 483.840 
>> 116.640 0.000 825.230
>>
>> Why ten numbers, I have 8 thraeds ! ? here for the last number for 
>> exemple, I had 825seconds that thread number 8 was at 230% of it's 
>> maximum ???
>
>
> The following post from Neil Brown has a 'better' explanation:
>
> http://marc.theaimsgroup.com/?l=linux-nfs&m=102824853219024&w=2
>
> James Pearson

Very well, It is clear to me now :-) !
however how can I zero these counters ?, restarting /etc/init.d/nfs 
doesn't do it, neither :
$ nfsstat -z
nfsstat: zeroing of nfs server statistics is not supported
nfsstat: zeroing of nfs client statistics is not supported

Is the only solution is to rebbot :-( ?


-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: interprtation of /proc/net/rpc/nfsd

[Steve Dickson]

Jehan PROCACCIA wrote:

> however how can I zero these counters ?, restarting /etc/init.d/nfs 
> doesn't do it, neither :
> $ nfsstat -z
> nfsstat: zeroing of nfs server statistics is not supported
> nfsstat: zeroing of nfs client statistics is not supported
>
Well... zeroing out the nfs stats was something I proposed a while ago
but got shoot down both internally and externally since zero out stats
is "not the Linux way".... so I've taken the patch out of both FC and RHEL
kernels, but I guess I forgot to take it out of the nfs-utils....

I could probably dig up the patch if your interested.....

> Is the only solution is to rebbot :-( ?

No. "rmmod nfs && modprobe nfs" also does the trick....

SteveD.


-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: interprtation of /proc/net/rpc/nfsd

[Jehan PROCACCIA]

Steve Dickson wrote:

> Jehan PROCACCIA wrote:
>
>> however how can I zero these counters ?, restarting /etc/init.d/nfs 
>> doesn't do it, neither :
>> $ nfsstat -z
>> nfsstat: zeroing of nfs server statistics is not supported
>> nfsstat: zeroing of nfs client statistics is not supported
>>
> Well... zeroing out the nfs stats was something I proposed a while ago
> but got shoot down both internally and externally since zero out stats
> is "not the Linux way".... so I've taken the patch out of both FC and 
> RHEL
> kernels, but I guess I forgot to take it out of the nfs-utils....
>
> I could probably dig up the patch if your interested.....

I was, but if it's not recommended ...., however your solution below 
(/etc/init.d/nfs stop; modprobe -r nfsd; /etc/init.d/nfs start) does 
perfecly the job of zeroing stats :-)

Thanks .

>
>> Is the only solution is to rebbot :-( ?
>
>
> No. "rmmod nfs && modprobe nfs" also does the trick....
>
> SteveD.




-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs