Process take long time to write to NFS filesystem.

[unix syzadmin]

Eugene,

Thanks for all your response.
I still wonder how the nfs writes go fine when their is enough free memory.

We did some dd tests writing and reading 16GB file to NFS filesystem in 64K
chunks.
>From vmstat we see a lot of memory was freed during read operation, but not
wirte.  Could this be the reason?


# time dd if=/dev/zero of=/zulu/staging/testfile bs=64k count=262144
262144+0 records in
262144+0 records out
17179869184 bytes (17 GB) copied, 846.982 seconds, 20.3 MB/s
real    14m6.990s
user    0m2.998s
sys     9m21.523s

# time dd if=/zulu/staging/testfile of=/dev/null bs=64k
262144+0 records in
262144+0 records out
17179869184 bytes (17 GB) copied, 779.67 seconds, 22.0 MB/s
real    12m59.678s
user    0m0.499s
sys     2m9.118s

******* vmstat capture during dd read test ********
********************************************************
procs -----------memory---------- ---swap-- -----io---- --system--
-----cpu------
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id
wa st
 0  0     84 222292 156132 15151496    0    0     0    18 1028  192  0  0
100  0  0
 0  0     84 222292 156136 15151496    0    0     0    42 1016  174  0  0
99  0  0
 0  0     84 222292 156136 15151496    0    0     0    25 1025  178  2  0
98  0  0

 0  1     84 14317360 156136 1088028    0    0     0     9 1032  180  0 40
52  8  0
 0  1     84 14281060 156136 1124412    0    0     0    34 1438  276  0  6
46 48  0
 1  1     84 14178340 156136 1227224    0    0     0     6 1931  427  2 18
36 44  0
 0  1     84 14069272 156136 1336580    0    0     0    14 1868  456  1 24
36 39  0
 0  1     84 13948188 156140 1457640    0    0     0    17 1918  430  0 20
37 43  0
 0  1     84 13821468 156140 1584260    0    0     0     9 2054  451  2 20
36 42  0
 0  1     84 13661860 156140 1743656    0    0     0     8 2228  524  0 27
35 37  0


Thanks.

On Sat, Jan 2, 2010 at 5:58 PM, Eugene Vilensky <evilensky at gmail.com> wrote:

> > Also what is the output of df for this particular volume?  What about
> > the storage aggregate?  For NetApp you need some free space inside the
> > volume, so that you are not stuck waiting for free space to be created
> > from deleted files (it takes time for blocks to be freed after a DEL),
> > and also plenty of free space at the aggregate level for new writes.
>
> And one last thing, make sure all best-practices are being followed
> for the back-end storage that resides behind the V-series NetApp.
> While the NetApp services the NFS, it is really the back end disks
> that are being ultimately written to, so make sure whether it is
> HDS/EMC/whatever, it is properly configured.  Misconfigured storage is
> unfortunately very easy to do.
>
> --
> redhat-list mailing list
> unsubscribe mailto:redhat-list-request at redhat.com?subject=unsubscribe
> https://www.redhat.com/mailman/listinfo/redhat-list
>