|
From: | Ian Rogers |
Subject: | [Gluster-devel] split-brain [was ping timeout] |
Date: | Thu, 25 Mar 2010 19:25:28 +0000 |
User-agent: | Mozilla/5.0 (Windows; U; Windows NT 5.1; en-GB; rv:1.8.1.23) Gecko/20090812 Thunderbird/2.0.0.23 Mnenhy/0.7.6.666 |
Having read all the previous posts I think there's some things we agree on wrt. split-brain
1. We wish Vikas would be given more time to finish http://www.gluster.com/community/documentation/index.php/Internals_of_Replicate :-)
2. The clients could do with some kind of optional quorum system to detect split-brain if sufficient sub-volumes are uncontactable (and not intentionally downed by the admin). The client could then go into read-only or totally-down mode depending on further options
3. ls -laR is just too slow as a method to re-sync large volumes. In the case where a sub-volume dissappears but the whole volume is not marked read-only (e.g. when a sub-volume is deliberately taken off-line so "sub quorum" is not flagged) the clients need some kind of dirty-file list so they can spawn off a thread to re-sync them quickly when the sub-volume(s) rejoin the client.
Are there any (more) rich companies out there who can buy a few extra support licenses to get this done? Or perhaps open up a "small FOSS donations" channel in pay-pal or something?... :-)
Ian
[Prev in Thread] | Current Thread | [Next in Thread] |