[Bugs] [Bug 1138970] file corruption during concurrent read/write

bugzilla at redhat.com bugzilla at redhat.com
Tue Oct 21 07:11:30 UTC 2014


https://bugzilla.redhat.com/show_bug.cgi?id=1138970



--- Comment #13 from Raghavendra G <rgowdapp at redhat.com> ---
Flipping the order of md-cache and write-behind xlators in xlator graph makes
the issue go away. This combined with the observation that file is never
corrupted on back-end points to caching gone bad in the layers above glusterfs
- either linux vfs or python File object libraries in this case. I tried to
verify by setting --direct-io-mode=yes during glusterfs mount. However, the
attached script hangs in an infinite loop while trying to read the file (except
for first few reads, reads are not sent to glusterfs at all).

So, there are two possible fixes to this problem:

1. Either have some communication b/w write-behind and md-cache to indicate
that stat is invalid when writes are cached by write-behind. The communication
can be an explicit one through xdata or an implicit one in the form of zeroed
out iatt etc.

2. Flip the order or write-behind and md-cache in xlator graph.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
Unsubscribe from this bug https://bugzilla.redhat.com/token.cgi?t=aMfOZll5aq&a=cc_unsubscribe


More information about the Bugs mailing list