Sorry for the delay -- here's what the volume looks like. It's pretty basic:
volume elkguests0-server
type protocol/server
option rpc-auth-allow-insecure on
option auth.addr./mnt/bulk.allow *
option auth.login.39da7ae0-8cee-4152-a612-674c48da544e.password 62a8ff81-cd3d-4872-a9b9-bad5da242f10
option auth.login./mnt/bulk.allow 39da7ae0-8cee-4152-a612-674c48da544e
option transport-type tcp
subvolumes /mnt/bulk
end-volume
type=2
count=6
status=1
sub_count=2
stripe_count=1
replica_count=2
version=12
transport-type=0
volume-id=7c50ebe2-337b-4bed-a53b-5fa706150b26
username=39da7ae0-8cee-4152-a612-674c48da544e
password=62a8ff81-cd3d-4872-a9b9-bad5da242f10
op-version=2
client-op-version=2
server.allow-insecure=on
nfs.disable=0
brick-0=fs-elkpo01.suntradingllc.com:-mnt-bulk
brick-1=fs-elkpo02.suntradingllc.com:-mnt-bulk
brick-2=fs-elkpo03.suntradingllc.com:-mnt-bulk
brick-3=fs-elkpo04.suntradingllc.com:-mnt-bulk
brick-4=fs-elkpo05.suntradingllc.com:-mnt-bulk
brick-5=fs-elkpo06.suntradingllc.com:-mnt-bulk
I'll try and dig up some client logs, but the last event was over a month ago.
From: "Vijay Bellur" <vbellur@xxxxxxxxxx>
To: "Nicholas Majeran" <nmajeran@xxxxxxxxxxxxxxxxx>, gluster-users@xxxxxxxxxxx, "Shyamsundar Ranganathan" <srangana@xxxxxxxxxx>
Sent: Monday, February 17, 2014 10:16:56 AM
Subject: Re: add-brick and fix-layout takes some VMs offline
On 02/13/2014 08:52 PM, Nicholas Majeran wrote:
> Hi there,
>
> We have a distributed-replicated volume hosting KVM guests running
> Gluster 3.4.1.
> We've grown from 1 x 2 -> 2 x 2 -> 3 x 2,but each time we've added nodes
> or run a fix layout,
> some of our guests go offline (or worse with error=continue they
> silently error).
> After the last addition we didn't even run fix-layout as the guests are
> becoming increasingly important.
Would it be possible to share the client log files and your volume
configuration?
>
> Those guests are currently are using a combination of FUSE and libgfapi.
> Is there a setting or group of settings we should use to ameliorate this
> problem?
> Is FUSE or libgfapi more forgiving when add-brick or fix-layout is run?
>
The behavior of FUSE or libgfapi should mostly be the same with either
add-brick or fix-layout.
Thanks,
Vijay
To: "Nicholas Majeran" <nmajeran@xxxxxxxxxxxxxxxxx>, gluster-users@xxxxxxxxxxx, "Shyamsundar Ranganathan" <srangana@xxxxxxxxxx>
Sent: Monday, February 17, 2014 10:16:56 AM
Subject: Re: add-brick and fix-layout takes some VMs offline
On 02/13/2014 08:52 PM, Nicholas Majeran wrote:
> Hi there,
>
> We have a distributed-replicated volume hosting KVM guests running
> Gluster 3.4.1.
> We've grown from 1 x 2 -> 2 x 2 -> 3 x 2,but each time we've added nodes
> or run a fix layout,
> some of our guests go offline (or worse with error=continue they
> silently error).
> After the last addition we didn't even run fix-layout as the guests are
> becoming increasingly important.
Would it be possible to share the client log files and your volume
configuration?
>
> Those guests are currently are using a combination of FUSE and libgfapi.
> Is there a setting or group of settings we should use to ameliorate this
> problem?
> Is FUSE or libgfapi more forgiving when add-brick or fix-layout is run?
>
The behavior of FUSE or libgfapi should mostly be the same with either
add-brick or fix-layout.
Thanks,
Vijay
_______________________________________________ Gluster-users mailing list Gluster-users@xxxxxxxxxxx http://supercolony.gluster.org/mailman/listinfo/gluster-users
