cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Brenn Oosterbaan" <boosterb...@schubergphilis.com>
Subject Review Request 24692: CLOUDSTACK-7345 patch to change dd command from 4M to 128k when using NFS
Date Thu, 14 Aug 2014 07:51:31 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/24692/
-----------------------------------------------------------

Review request for cloudstack, daan Hoogland, Joris van Lieshout, and Sanjay Tripathi.


Bugs: CLOUDSTACK-7345
    https://issues.apache.org/jira/browse/CLOUDSTACK-7345


Repository: cloudstack-git


Description
-------

CLOUDSTACK-7345 patch to change dd command from 4M to 128k when using NFS


Diffs
-----

  scripts/vm/hypervisor/xenserver/vmopsSnapshot 85444dc 

Diff: https://reviews.apache.org/r/24692/diff/


Testing
-------

Starting multiple dd commands with bs=4M on a single hypervisor causes nfs server timed out
messages in /var/log/kern.log and causes the dd process to crash.
With bs=4M NFS debugging shows 64 seperate 64k reads are done, after which the inode is updated
64 times (access time) and 32 128k writes are done to secondary storage.
NFS debugging showed the 'nfs server timed out' messages usually occured during the 64 inode
updates (of the same file) from the read process.

Running multiple dd commands with bs=128k on a single hypervisor does not cause nfs server
timed out messages in /var/log/kern.log.
With bs=128k NFS debugging shows 2 seperate 64k reads are done (and indoe updates) and 1 128k
write is done.

We are running this change in production. The nfs server timed out messages are gone and not
a single snapshot process has broken (previosuly this happened every night).


Thanks,

Brenn Oosterbaan


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message