[Openstack] [Swift] block I/O all disks
Mark Kirkwood
mark.kirkwood at catalyst.net.nz
Wed Nov 11 05:53:18 UTC 2015
On 11/11/15 00:59, Heiko Krämer wrote:
> Hi Mark,
>
> thanks for your response.
> I changed the dirty page settings again and i have a logging server
> which gets the syslog and swift logs.
>
>
>
> But this messages will not help to find the problem. They are stale
> processes and it seems the start partition is the SSD Raid1 which caused
> in a cascade overall SATA HDD's.
>
(I'm gonna bottom post replies fyi...)
Ah, good sleuthing! So I wonder if your SSD are the problem - what make
and model are they? Some do *not* like being RAID'ed. We pretty much
will only use Intel DC37xx series as they seem to avoid all the nasty
edge case behaviours that most other models exhibit (stalls, sudden
catastrophic failues)...
I'm still not 100% clear on your setup, are you using the SSD for
account and container servers? or is all your Swift stuff on the SATA
drives? *If* you are using the SSD for account + container you might be
better off not RAIDing them.
regards
mark
More information about the Openstack
mailing list