[Bugs] [Bug 1416321] read error: Input/ output error but not in brain split stat while file on bricks has different sizes

[bugzilla at redhat.com]

https://bugzilla.redhat.com/show_bug.cgi?id=1416321

Pranith Kumar K <pkarampu at redhat.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
              Flags|                            |needinfo?(likunbyl at qq.com)



--- Comment #22 from Pranith Kumar K <pkarampu at redhat.com> ---
(In reply to Pranith Kumar K from comment #21)
> (In reply to likunbyl from comment #19)
> > (In reply to Ravishankar N from comment #18)
> > > 
> > > I thought I already answered that. AFR==client side.
> > 
> > Do I need to close this case ?
> 
> I think we don't yet have any theory as to why this might have happened, so
> while upgrading to a supported version is recommended, it doesn't give us
> any confidence about if the problem we experienced in the existing setup is
> already solved or not in the newer version.
> 
> We used to see this kind of problem when any of the disks crash where the
> sizes would mismatch but there wouldn't be any extended attributes to tell
> that this happened.

Just to want to clarify that if a file ends up in the state you mentioned in
the bz, the file will be automatically healed in versions > 3.6.x so if you
upgrade to 3.8.x then you won't get EIO when the file ends up in this state. So
upgrade is recommended.

What I am curious to know is how did the file end up in this state. We used to
see this kind of issue only when there is a disk/filesystem crash or ungraceful
shutdown of the machine.

Did any of the above happen on the brick servers where the first brick is
present?

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are the assignee for the bug.