[PATCH] nfs-utils: mountd: empty export list should not be treated as a failure

[PATCH] nfs-utils: mountd: empty export list should not be treated as a failure

[Harshula]

Hi Steve,

This patch is against:
git://git.linux-nfs.org/projects/steved/nfs-utils.git

In mountd, if get_exportlist() (utils/mountd/mountd.c) returns NULL it
should not be considered a failure. It just means that there are no
exports on the system.

The practical problem with the current code is that a showmount -e
results in a syslog message from mountd that looks like:

rpc.mountd: export request from 10.250.100.2 failed.


References: SGI: PV977213
Reviewed-by: Greg Banks <gnb@sgi.com>
Signed-off-by: Harshula Jayasuriya <harshula@sgi.com>
---

 utils/mountd/mountd.c |   11 ++++-------
 1 file changed, 4 insertions(+), 7 deletions(-)

--- a/utils/mountd/mountd.c
+++ b/utils/mountd/mountd.c
@@ -254,10 +254,8 @@ mount_export_1_svc(struct svc_req *rqstp, void
*argp, exports *resp)
        struct sockaddr_in *addr =
                (struct sockaddr_in *) svc_getcaller(rqstp->rq_xprt);
 
-       if ((*resp = get_exportlist()) == NULL)
-               xlog(L_WARNING, "export request from %s failed.",
-                       inet_ntoa(addr->sin_addr));
-
+       xlog(D_CALL, "EXPORT1 request from %s.",
inet_ntoa(addr->sin_addr));
+       *resp = get_exportlist();
        return 1;
 }
 
@@ -267,9 +265,8 @@ mount_exportall_1_svc(struct svc_req *rqstp, void
*argp, exports *resp)
        struct sockaddr_in *addr =
                (struct sockaddr_in *) svc_getcaller(rqstp->rq_xprt);
 
-       if ((*resp = get_exportlist()) == NULL)
-               xlog(L_WARNING, "exportall request from %s failed.",
-                       inet_ntoa(addr->sin_addr));
+       xlog(D_CALL, "EXPORTALL1 request from %s.",
inet_ntoa(addr->sin_addr));
+       *resp = get_exportlist();
        return 1;
 }
 

cya,
#

nfs performance dropped when using 128 MB RAM

[Sagar Borikar]

[-- Attachment #1: Type: text/plain, Size: 2747 bytes --]

 Hello Folks,
 
I am not sure if this is the right place for this question. If not, I am
sorry and request you to direct me to right mailing list.
 
Currently I am working on a product which is derived from its ancestor
with more or less similar feature set and few additions. Other than
memory size, there are not hardware changes. The newer one is having 128
MB RAM while the ancestor was having 512 MB of RAM. The functionality
carried forward is also similar except for nfs. Both have gigabit
Ethernet interfaces
                 I am finding that nfs write performance has drastically
reduced say 1 MB/s or even it stalls for large file transfers to the
box. On the contrary it works fine with 512 MB of RAM. The CPU
utilization by nfsd as well as rpc.statd / rpc.mountd is also non
existent. Rather nfsd most of the time is sleeping for the data.  While
trying to analyse through ethereal, I observe that the write requests by
clients to the the server are periodic say around 1.5-2 minutes
interval. Till then the tcpdump captures that the handshake is going on
only for ack messages for wait requests. After then suddenly the burst
write transfer starts and within 4-5 seconds when the buffers are full
and no free memory is available, again it repeats the same cycle. When
poking in the meminfo stats, I found that there are no dirty pages and
write back buffers available before write bursts. As soon as the buffers
are available, nfs sends requests for further data and meminfo stats are
as it is. The free memory which is LowFree memory used by kernel data
structures in our case as High free memory is zero, is very less when
the buffers are not cleaned up. Once the burst data transfer finishes,
cached memory equal to that of previous buffered gets cleared as
expected. But Inactive memory is still high. I am attaching the tcpdump
logs with this mail. I am using rsize / wsize as 64k. Tried various
combinations of these sizes while mounting the nfs clients but it didn't
help.
I tried to check whether this problem is system wide but unfortunately
CIFS and FTP writes are pretty fast. so only nfs performance is affected
due to RAM size reduction. We have enabled socket buffer recycling in
Ethernet driver which block 600k of data for recycling. Even just to get
confirmation, I disabled the skb  recycling, even then the result was
same. I am not seeing this kind of behavior in other workstations with
128 MB of RAM and mips architecture.
Can anyone let me know, why the write requests are buffered and
periodically they get cleared? 
Also does nfs stack, requires special buffer availability for succeeding
write requests? 
 
Thanks and Regards,
Sagar Borikar 
PMC sierra
 
 

[-- Attachment #2: tcpdump.log --]
[-- Type: application/octet-stream, Size: 7937 bytes --]


Mstening on eth0, link-type EN10MB (Ethernet), capture size 96 bytes
11:12:35.674988 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8161140 win 2896 <nop,nop,timestamp 18958133 515936722>
11:12:35.675005 IP 172.16.48.52.166575521 > 172.16.48.130.nfs: 2896 proc-2379056868
11:12:35.675013 IP 172.16.48.52.415833216 > 172.16.48.130.nfs: 1448 proc-3037336800
11:12:35.675394 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8164036 win 2896 <nop,nop,timestamp 18958133 515936722>
11:12:35.675411 IP 172.16.48.52.2455467960 > 172.16.48.130.nfs: 1448 proc-3812267133
11:12:35.675419 IP 172.16.48.52.26272008 > 172.16.48.130.nfs: 1448 proc-3094064178
11:12:35.675636 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8166932 win 2896 <nop,nop,timestamp 18958133 515936723>
11:12:35.675652 IP 172.16.48.52.817467000 > 172.16.48.130.nfs: 1448 proc-1456776907
11:12:35.675660 IP 172.16.48.52.3578191046 > 172.16.48.130.nfs: 1448 proc-1902769939
11:12:35.675924 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8169828 win 2896 <nop,nop,timestamp 18958133 515936723>
11:12:35.675940 IP 172.16.48.52.3721596643 > 172.16.48.130.nfs: 1448 proc-974205282
11:12:35.675947 IP 172.16.48.52.2604580239 > 172.16.48.130.nfs: 1448 proc-1590512167
11:12:35.676169 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8172724 win 2896 <nop,nop,timestamp 18958133 515936723>
11:12:35.676185 IP 172.16.48.52.4278242473 > 172.16.48.130.nfs: 1448 proc-3126856275
11:12:35.676193 IP 172.16.48.52.2419163184 > 172.16.48.130.nfs: 1448 proc-1815905253
11:12:35.676877 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8175620 win 2896 <nop,nop,timestamp 18958134 515936724>
11:12:35.676896 IP 172.16.48.52.528592223 > 172.16.48.130.nfs: 1448 proc-3790581237
11:12:35.676904 IP 172.16.48.52.674712662 > 172.16.48.130.nfs: 1448 proc-1923312982
11:12:35.676913 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8178516 win 2896 <nop,nop,timestamp 18958134 515936724>
11:12:35.676928 IP 172.16.48.52.2019061448 > 172.16.48.130.nfs: 1448 proc-3786701257
11:12:35.676935 IP 172.16.48.52.3938107219 > 172.16.48.130.nfs: 1448 proc-477133596
11:12:35.677420 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8181412 win 2896 <nop,nop,timestamp 18958134 515936725>
11:12:35.677439 IP 172.16.48.52.3303078522 > 172.16.48.130.nfs: 1448 proc-416174347
11:12:35.677447 IP 172.16.48.52.992055196 > 172.16.48.130.nfs: 1448 proc-3167533369
11:12:35.677659 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8184308 win 2896 <nop,nop,timestamp 18958134 515936725>
11:12:35.677675 IP 172.16.48.52.3782641007 > 172.16.48.130.nfs: 1448 proc-3744596530
11:12:35.677683 IP 172.16.48.52.2800514754 > 172.16.48.130.nfs: 1448 proc-1632170636
11:12:35.677956 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8187204 win 2896 <nop,nop,timestamp 18958134 515936725>
11:12:35.677972 IP 172.16.48.52.2920219751 > 172.16.48.130.nfs: 1448 proc-167343746
11:12:35.677980 IP 172.16.48.52.1216585728 > 172.16.48.130.nfs: 1448 proc-2784742365
11:12:35.678228 IP 172.16.48.130.nfs > 172.16.48.52.2116931643: reply ok 140
11:12:35.678245 IP 172.16.48.52.798423794 > 172.16.48.130.nfs: 1448 proc-1298995762
11:12:35.678863 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8194444 win 2896 <nop,nop,timestamp 18958134 515936725>
11:12:35.678880 IP 172.16.48.52.1705633946 > 172.16.48.130.nfs: 2896 proc-2178937653
11:12:35.678888 IP 172.16.48.52.865072077 > 172.16.48.130.nfs: 2896 proc-3935072703
11:12:35.679395 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8197340 win 2896 <nop,nop,timestamp 18958134 515936727>
11:12:35.679412 IP 172.16.48.52.871577439 > 172.16.48.130.nfs: 2896 proc-3341172484
11:12:35.679640 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8200236 win 2896 <nop,nop,timestamp 18958134 515936727>
11:12:35.679655 IP 172.16.48.52.1517075806 > 172.16.48.130.nfs: 2896 proc-2824690568
11:12:35.679925 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8203132 win 2896 <nop,nop,timestamp 18958134 515936727>
11:12:35.679942 IP 172.16.48.52.2451863779 > 172.16.48.130.nfs: 2896 proc-2104363821
11:12:35.680173 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8206028 win 2896 <nop,nop,timestamp 18958134 515936727>
11:12:35.680188 IP 172.16.48.52.2425756098 > 172.16.48.130.nfs: 2896 proc-4117019690
11:12:35.680777 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8208924 win 2896 <nop,nop,timestamp 18958135 515936728>
11:12:35.680794 IP 172.16.48.52.2944274149 > 172.16.48.130.nfs: 2896 proc-2117355039
11:12:35.680798 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8211820 win 2896 <nop,nop,timestamp 18958135 515936728>
11:12:35.680809 IP 172.16.48.52.1728566045 > 172.16.48.130.nfs: 2896 proc-2220972509
11:12:35.681317 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8214716 win 2896 <nop,nop,timestamp 18958135 515936729>
11:12:35.681333 IP 172.16.48.52.1125632345 > 172.16.48.130.nfs: 2896 proc-1984113115
11:12:35.681554 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8217612 win 2896 <nop,nop,timestamp 18958135 515936729>
11:12:35.681570 IP 172.16.48.52.2052363586 > 172.16.48.130.nfs: 2896 proc-3058647012
11:12:35.681851 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8220508 win 2896 <nop,nop,timestamp 18958135 515936729>
11:12:35.681867 IP 172.16.48.52.3169922154 > 172.16.48.130.nfs: 2896 proc-1693529181
11:12:35.682000 IP 172.16.48.130.nfs > 172.16.48.52.2133708859: reply ok 140
11:12:35.682601 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8226300 win 2896 <nop,nop,timestamp 18958135 515936729>
11:12:35.682617 IP 172.16.48.52.206440654 > 172.16.48.130.nfs: 2896 proc-2400289967
11:12:35.682625 IP 172.16.48.52.2681648301 > 172.16.48.130.nfs: 2896 proc-969898632
11:12:35.683132 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8229196 win 2896 <nop,nop,timestamp 18958135 515936730>
11:12:35.683149 IP 172.16.48.52.166182463 > 172.16.48.130.nfs: 2896 proc-4137940490
11:12:35.683379 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8232092 win 2896 <nop,nop,timestamp 18958135 515936730>
11:12:35.683395 IP 172.16.48.52.870457670 > 172.16.48.130.nfs: 2896 proc-460917824
11:12:35.683663 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8234988 win 2896 <nop,nop,timestamp 18958135 515936731>
11:12:35.683678 IP 172.16.48.52.921558232 > 172.16.48.130.nfs: 2896 proc-1260781838
11:12:35.683905 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8237884 win 2896 <nop,nop,timestamp 18958135 515936731>
11:12:35.683922 IP 172.16.48.52.2119784769 > 172.16.48.130.nfs: 2896 proc-4255658928
11:12:35.684570 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8240780 win 2896 <nop,nop,timestamp 18958136 515936731>
11:12:35.684586 IP 172.16.48.52.2604214023 > 172.16.48.130.nfs: 2896 proc-4039840750
11:12:35.684594 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8243676 win 2896 <nop,nop,timestamp 18958136 515936732>
11:12:35.684607 IP 172.16.48.52.2031761012 > 172.16.48.130.nfs: 2896 proc-4195605192
11:12:35.685115 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 8246572 win 2896 <nop,nop,timestamp 18958136 515936732>
11:12:35.685133 IP 172.16.48.52.3846229746 > 172.16.48.130.nfs: 2896 proc-2380498847
11:12:44.855773 IP 172.16.48.52.884 > 172.16.48.130.nfs: . ack 3396395833 win 501 <nop,nop,timestamp 515945904 18930424>
11:12:44.856003 IP 172.16.48.130.nfs > 172.16.48.52.884: . ack 1 win 0 <nop,nop,timestamp 18960429 515619303>
11:12:45.227822 IP 172.16.48.52.879 > 172.16.48.130.nfs: . ack 4074022534 win 729 <nop,nop,timestamp 515946276 18959710>
11:12:45.228049 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 1 win 0 <nop,nop,timestamp 18960522 515939986>
11:12:51.128765 IP 172.16.48.52.882 > 172.16.48.130.nfs: . ack 3871614401 win 501 <nop,nop,timestamp 515952177 18936137>
11:12:51.128972 IP 172.16.48.130.nfs > 172.16.48.52.882: . ack 1 win 0 <nop,nop,timestamp 18961997 515745571>
11:12:51.723781 IP 172.16.48.52.879 > 172.16.48.130.nfs: . ack 1 win 729 <nop,nop,timestamp 515952772 18960522>
11:12:51.723948 IP 172.16.48.130.nfs > 172.16.48.52.879: . ack 1 win 0 <nop,nop,timestamp 18962146 515939986>

Re: nfs performance dropped when using 128 MB RAM

[J. Bruce Fields]

On Tue, Feb 12, 2008 at 12:33:41AM -0800, Sagar Borikar wrote:
>  Hello Folks,
>  
> I am not sure if this is the right place for this question. If not, I am
> sorry and request you to direct me to right mailing list.

This is the right list, but you're replying to an unrelated message,
which confuses those of us with threaded mail readers--no big deal, just
please be more careful about that next time.

> Currently I am working on a product which is derived from its ancestor
> with more or less similar feature set and few additions. Other than
> memory size, there are not hardware changes. The newer one is having 128
> MB RAM while the ancestor was having 512 MB of RAM. The functionality
> carried forward is also similar except for nfs. Both have gigabit
> Ethernet interfaces

This is a server, not a client?  What kernel are you using?

--b.

>                  I am finding that nfs write performance has drastically
> reduced say 1 MB/s or even it stalls for large file transfers to the
> box. On the contrary it works fine with 512 MB of RAM. The CPU
> utilization by nfsd as well as rpc.statd / rpc.mountd is also non
> existent. Rather nfsd most of the time is sleeping for the data.  While
> trying to analyse through ethereal, I observe that the write requests by
> clients to the the server are periodic say around 1.5-2 minutes
> interval. Till then the tcpdump captures that the handshake is going on
> only for ack messages for wait requests. After then suddenly the burst
> write transfer starts and within 4-5 seconds when the buffers are full
> and no free memory is available, again it repeats the same cycle. When
> poking in the meminfo stats, I found that there are no dirty pages and
> write back buffers available before write bursts. As soon as the buffers
> are available, nfs sends requests for further data and meminfo stats are
> as it is. The free memory which is LowFree memory used by kernel data
> structures in our case as High free memory is zero, is very less when
> the buffers are not cleaned up. Once the burst data transfer finishes,
> cached memory equal to that of previous buffered gets cleared as
> expected. But Inactive memory is still high. I am attaching the tcpdump
> logs with this mail. I am using rsize / wsize as 64k. Tried various
> combinations of these sizes while mounting the nfs clients but it didn't
> help.
> I tried to check whether this problem is system wide but unfortunately
> CIFS and FTP writes are pretty fast. so only nfs performance is affected
> due to RAM size reduction. We have enabled socket buffer recycling in
> Ethernet driver which block 600k of data for recycling. Even just to get
> confirmation, I disabled the skb  recycling, even then the result was
> same. I am not seeing this kind of behavior in other workstations with
> 128 MB of RAM and mips architecture.
> Can anyone let me know, why the write requests are buffered and
> periodically they get cleared? 
> Also does nfs stack, requires special buffer availability for succeeding
> write requests? 
>  
> Thanks and Regards,
> Sagar Borikar 
> PMC sierra
>  
>  

Re: [PATCH] nfs-utils: mountd: empty export list should not be treated as a failure

[Steve Dickson]

Harshula wrote:
> In mountd, if get_exportlist() (utils/mountd/mountd.c) returns NULL it
> should not be considered a failure. It just means that there are no
> exports on the system.
> 
> The practical problem with the current code is that a showmount -e
> results in a syslog message from mountd that looks like:
> 
> rpc.mountd: export request from 10.250.100.2 failed.
Committed... thanks!

steved.

RE: nfs performance dropped when using 128 MB RAM

[Sagar Borikar]

On Tue, Feb 12, 2008 at 12:33:41AM -0800, Sagar Borikar wrote:
>  Hello Folks,
> 
> I am not sure if this is the right place for this question. If not, I am
> sorry and request you to direct me to right mailing list.

This is the right list, but you're replying to an unrelated message,
which confuses those of us with threaded mail readers--no big deal, just
please be more careful about that next time.
 

> Currently I am working on a product which is derived from its ancestor
> with more or less similar feature set and few additions. Other than
> memory size, there are not hardware changes. The newer one is having 128
> MB RAM while the ancestor was having 512 MB of RAM. The functionality
> carried forward is also similar except for nfs. Both have gigabit
> Ethernet interfaces

This is a server, not a client?  What kernel are you using?

<Sagar> That's right. It is server and kernel is 2.6.18. I would appreciate any pointers / directions with this regard.

Thanks


>                  I am finding that nfs write performance has drastically
> reduced say 1 MB/s or even it stalls for large file transfers to the
> box. On the contrary it works fine with 512 MB of RAM. The CPU
> utilization by nfsd as well as rpc.statd / rpc.mountd is also non
> existent. Rather nfsd most of the time is sleeping for the data.  While
> trying to analyse through ethereal, I observe that the write requests by
> clients to the the server are periodic say around 1.5-2 minutes
> interval. Till then the tcpdump captures that the handshake is going on
> only for ack messages for wait requests. After then suddenly the burst
> write transfer starts and within 4-5 seconds when the buffers are full
> and no free memory is available, again it repeats the same cycle. When
> poking in the meminfo stats, I found that there are no dirty pages and
> write back buffers available before write bursts. As soon as the buffers
> are available, nfs sends requests for further data and meminfo stats are
> as it is. The free memory which is LowFree memory used by kernel data
> structures in our case as High free memory is zero, is very less when
> the buffers are not cleaned up. Once the burst data transfer finishes,
> cached memory equal to that of previous buffered gets cleared as
> expected. But Inactive memory is still high. I am attaching the tcpdump
> logs with this mail. I am using rsize / wsize as 64k. Tried various
> combinations of these sizes while mounting the nfs clients but it didn't
> help.
> I tried to check whether this problem is system wide but unfortunately
> CIFS and FTP writes are pretty fast. so only nfs performance is affected
> due to RAM size reduction. We have enabled socket buffer recycling in
> Ethernet driver which block 600k of data for recycling. Even just to get
> confirmation, I disabled the skb  recycling, even then the result was
> same. I am not seeing this kind of behavior in other workstations with
> 128 MB of RAM and mips architecture.
> Can anyone let me know, why the write requests are buffered and
> periodically they get cleared?
> Also does nfs stack, requires special buffer availability for succeeding
> write requests?
> 
> Thanks and Regards,
> Sagar Borikar
> PMC sierra
> 
> 

RE: nfs performance dropped when using 128 MB RAM

[Sagar Borikar]

 
Folks,

Ok, I was able to crack the issue but in some odd way. I found that
pdflush was not able to flush the buffers in time because of which the
next write requests were stalled at the client side. I did following
changes to improve the performance. 

1. Changed the number of pdflush threads from default 2 to 4
2. Changed the CACHESIZE from 1024 to 8192 in fs/nfsd/nfscache.c
3. Changed SVC block size from 32k to 64k in include/linux/nfsd/const.h

After these changes, as expected I see much better performance but then
there is overshoot of at the start with less write size and later it
decreases. I captured the iozonelinux figures and they are as below

Excel output is below:
 
"Writer report"
        "4"  "8"  "16"  "32"  "64"
"65536"   1006667  1414856  1437128  1606784  1580778
"131072"   801006  1427892  1232945  1191835  1232504
"262144"   8555  16856  10093  16721  12607
"524288"   8136  13017  8419  11485  16160
"1048576"   6546  9635  10923  10611  10995
 
"Re-writer report"
        "4"  "8"  "16"  "32"  "64"
"65536"   1323668  1512831  1645809  1757521  1792708
"131072"   1119661  1299879  1367497  1423008  1427222
"262144"   3290  4173  3494  4158  4969
"524288"   3268  3398  3200  2899  3570
"1048576"   3126  3183  3646  3231  3890
 
"Reader report"
        "4"  "8"  "16"  "32"  "64"
"65536"   10790  11167  11123  10842  9409
"131072"   9270  10640  9084  9480  10958
"262144"   10136  11011  10881  8924  8911
"524288"   10947  9303  8567  9290  10853
"1048576"   10290  9499  10956  11156  11063
 
"Re-Reader report"
        "4"  "8"  "16"  "32"  "64"
"65536"   1770290  591736  686075  240800  236228
"131072"   753296  1219558  858200  1142548  1080169
"262144"   611654  308045  521012  403101  854802
"524288"   10834  9075  10613  9219  2725002
"1048576"   1410151  2341711  2635044  2869093  3004842
 
"Writer CPU utilization report (Zero values should be ignored)"
        "4"  "8"  "16"  "32"  "64"
"65536"     0.79   0.44   0.71   0.56   0.57
"131072"     0.75   0.78   0.57   0.67   0.72
"262144"     0.61   0.55   0.56   0.61   0.56
"524288"     0.68   0.58   0.59   0.26   0.59
"1048576"     0.61   0.68   0.67   0.68   0.68
 
"Re-writer CPU utilization report (Zero values should be ignored)"
        "4"  "8"  "16"  "32"  "64"
"65536"     0.70   0.66   0.59   0.52   0.70
"131072"     0.29   0.24   0.20   0.21   0.20
"262144"     0.25   0.22   0.19   0.20   0.20
"524288"     0.26   0.25   0.22   0.20   0.21
"1048576"     0.27   0.23   0.21   0.20   0.20
 
"Reader CPU utilization report (Zero values should be ignored)"
        "4"  "8"  "16"  "32"  "64"
"65536"     1.17   0.92   1.14   0.93   0.75
"131072"     1.00   1.17   0.80   0.79   1.07
"262144"     1.17   1.11   0.88   0.73   0.77
"524288"     1.19   0.90   0.77   0.78   0.84
"1048576"     1.13   0.91   0.99   0.94   0.87
 
"Re-Reader CPU utilization report (Zero values should be ignored)"
        "4"  "8"  "16"  "32"  "64"
"65536"    97.38  27.73  28.08  10.25   8.97
"131072"    41.15  57.14  33.57  43.21  44.13
"262144"    33.04  14.79  22.96  15.20  32.82
"524288"     1.15   0.81   0.93   0.75  93.97
"1048576"    71.14  99.75  99.18  99.22  99.61



I am surprised to see that write performance is better than read and
more interestingly re-write performance is degraded compared to write
performance.
Any cluses on this anomaly?

Thanks in advance

Sagar

-----Original Message-----
From: linux-nfs-owner@vger.kernel.org
[mailto:linux-nfs-owner@vger.kernel.org] On Behalf Of Sagar Borikar
Sent: Tuesday, February 12, 2008 2:04 PM
To: linux-nfs@vger.kernel.org
Subject: nfs performance dropped when using 128 MB RAM

 Hello Folks,
 
I am not sure if this is the right place for this question. If not, I am
sorry and request you to direct me to right mailing list.
 
Currently I am working on a product which is derived from its ancestor
with more or less similar feature set and few additions. Other than
memory size, there are not hardware changes. The newer one is having 128
MB RAM while the ancestor was having 512 MB of RAM. The functionality
carried forward is also similar except for nfs. Both have gigabit
Ethernet interfaces
                 I am finding that nfs write performance has drastically
reduced say 1 MB/s or even it stalls for large file transfers to the
box. On the contrary it works fine with 512 MB of RAM. The CPU
utilization by nfsd as well as rpc.statd / rpc.mountd is also non
existent. Rather nfsd most of the time is sleeping for the data.  While
trying to analyse through ethereal, I observe that the write requests by
clients to the the server are periodic say around 1.5-2 minutes
interval. Till then the tcpdump captures that the handshake is going on
only for ack messages for wait requests. After then suddenly the burst
write transfer starts and within 4-5 seconds when the buffers are full
and no free memory is available, again it repeats the same cycle. When
poking in the meminfo stats, I found that there are no dirty pages and
write back buffers available before write bursts. As soon as the buffers
are available, nfs sends requests for further data and meminfo stats are
as it is. The free memory which is LowFree memory used by kernel data
structures in our case as High free memory is zero, is very less when
the buffers are not cleaned up. Once the burst data transfer finishes,
cached memory equal to that of previous buffered gets cleared as
expected. But Inactive memory is still high. I am attaching the tcpdump
logs with this mail. I am using rsize / wsize as 64k. Tried various
combinations of these sizes while mounting the nfs clients but it didn't
help.
I tried to check whether this problem is system wide but unfortunately
CIFS and FTP writes are pretty fast. so only nfs performance is affected
due to RAM size reduction. We have enabled socket buffer recycling in
Ethernet driver which block 600k of data for recycling. Even just to get
confirmation, I disabled the skb  recycling, even then the result was
same. I am not seeing this kind of behavior in other workstations with
128 MB of RAM and mips architecture.
Can anyone let me know, why the write requests are buffered and
periodically they get cleared? 
Also does nfs stack, requires special buffer availability for succeeding
write requests? 
 
Thanks and Regards,
Sagar Borikar
PMC sierra