Brenn Oosterbaan created CLOUDSTACK-7345:
--------------------------------------------
Summary: Copy snapshot command can cause NFS timeouts, breaking
the snapshot process.
Key: CLOUDSTACK-7345
URL: https://issues.apache.org/jira/browse/CLOUDSTACK-7345
Project: CloudStack
Issue Type: Bug
Security Level: Public (Anyone can view this level - this is the default.)
Components: Snapshot, XenServer
Affects Versions: 4.2.0, 4.1.0, 4.0.2, 4.0.1, 4.0.0
Reporter: Brenn Oosterbaan
When multiple snapshots are being transferred from NFS primary to secondary
storage at the same time, NFS timeouts on the primary storage can occur.
This then breaks most of the snapshot processes, resulting in corrupted (half
completed) snapshots.
The DD command used uses 'bs=4M', which is then chopped into 64 64k reads since
the primary storage is mounted with a blocksize of 64k, and written to the
secondary storage as 32 128k blocks (since secondary is mounted with blocksize
128k).
We created a patch changing the DD command to use 'bs=128k' (unless iSCSI is
used in which case it stays at bs=4M).
Snapshot processes no longer break, and no negative performance impact.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
