[Checkins] SVN: gocept.zeoraid/trunk/doc/OPERATION.txt Re #484727: Document the issues with recovery in a multi-ZEORaid setup.
Christian Theune
ct at gocept.com
Thu Nov 19 02:51:00 EST 2009
Log message for revision 105855:
Re #484727: Document the issues with recovery in a multi-ZEORaid setup.
Changed:
U gocept.zeoraid/trunk/doc/OPERATION.txt
-=-
Modified: gocept.zeoraid/trunk/doc/OPERATION.txt
===================================================================
--- gocept.zeoraid/trunk/doc/OPERATION.txt 2009-11-19 07:50:26 UTC (rev 105854)
+++ gocept.zeoraid/trunk/doc/OPERATION.txt 2009-11-19 07:50:59 UTC (rev 105855)
@@ -2,6 +2,7 @@
Operational notes
=================
+
Packing
=======
@@ -42,3 +43,22 @@
server becomes available. In case you want to start without the ZEO server
being available, you need to configure the option `wait false` into your
corresponding. zeoclient section.
+
+
+Recovery in Multi-ZEORaid setups
+================================
+
+If multiple ZEORaid servers are used in parallel and a storage has failed then
+recovery needs some extra steps:
+
+- Stop all ZEORaid servers except one
+
+- Recover the failed storage on the remaining ZEORaid server
+
+- Restart the redundant ZEORaid servers
+
+Note: Not following these steps won't cause a core meltdown, however, without
+following those steps you will not be able to successfully recover a failed
+storage under load. There is a feature wish recorded in launchpad that will
+implement a behaviour which will allow to keep the redundant ZEORaid servers
+running.
More information about the checkins
mailing list