Bug 764497 (GLUSTER-2765) - geo-replication should have mercy on brick failure
Summary: geo-replication should have mercy on brick failure
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: GLUSTER-2765
Product: GlusterFS
Classification: Community
Component: geo-replication
Version: mainline
Hardware: x86_64
OS: Linux
medium
low
Target Milestone: ---
Assignee: kaushik
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2011-04-15 06:45 UTC by Csaba Henk
Modified: 2011-04-25 12:22 UTC (History)
1 user (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed:
Regression: ---
Mount Type: ---
Documentation: ---
CRM:
Verified Versions:


Attachments (Terms of Use)

Description Csaba Henk 2011-04-15 06:45:56 UTC
If brick goes down and the file it provides vanish, gsyncd would just delete them from slave as well.

That's quite unacceptable. So if this happens, force gsyncd's gluster client to exit, making this way geo-replication defunct. When the user puts back the brick with replace-brick, geo-replication session will be deleted and when user knows that things went back to normal state (including a manual sync-back of files of the brick gone), then he can start it again.

Comment 1 Anand Avati 2011-04-15 07:58:36 UTC
PATCH: http://patches.gluster.com/patch/6892 in master (glusterd / geo-replication: have gsync's glusterfs client use assert-no-child-down for dht volume)

Comment 2 Anand Avati 2011-04-15 07:58:44 UTC
PATCH: http://patches.gluster.com/patch/6894 in master (DHT: Make assert-no-child-down a boolean option)

Comment 3 Anand Avati 2011-04-19 06:29:43 UTC
PATCH: http://patches.gluster.com/patch/6944 in master (mgmt/glusterd: do not allow replace-brick operations when geo-rep sessions are active on this volume.)

Comment 4 Csaba Henk 2011-04-25 09:22:06 UTC
The procedure for brick restoration using geo-replication is described here:

https://gist.github.com/e87ebea373bb67cf52b1

please use this as reference for verification.


Note You need to log in before you can comment on or make changes to this bug.