I’m currently testing the last dev version in cluster and projections go into Preparing state forever when the master dies…
Repro steps:
start a 3 node cluster with projections. Projections are running.
Kill the master. All projections go into Preparing step…
No other operation (restarting the node etc…) can make projections go beyond preparing step…
Is this a known issue with latest dev commits, or am I doing it wrong ?
cheers,
jeremie / thinkbeforecoding
Small update…
If I bring the whole cluster down and restart then two nodes, projections are running again…
So the projection is not totally dead, but need a fresh start.
Here, the parameters I use to start the cluster:
EventStore.ClusterNode.exe -ClusterSize 3 -GossipSeed 127.0.0.1:2114 -inthttpport 2112 -exthttpport 2113 -inttcpport 1112 -exttcpport 1113 -httpprefixes http://127.0.0.1:2113/ -runprojections all -discoverviadns false --db Events1
EventStore.ClusterNode.exe -ClusterSize 3 -GossipSeed 127.0.0.1:2112 -inthttpport 2114 -exthttpport 2115 -inttcpport 1114 -exttcpport 1115 -httpprefixes http://127.0.0.1:2115/ -runprojections all -discoverviadns false --db Events2
EventStore.ClusterNode.exe -ClusterSize 3 -GossipSeed 127.0.0.1:2112 -inthttpport 2116 -exthttpport 2117 -inttcpport 1116 -exttcpport 1117 -httpprefixes http://127.0.0.1:2117/ -runprojections all -discoverviadns false --db Events3
Mh I have these lines in the log of the node that is elected as Master after the first one has been killed:
[08076,15,13:18:24.100] _eventReaders is not empty after all the projections have been killed
[08076,16,13:18:24.100] _eventReaders is not empty after all the projections have been killed
[08076,17,13:18:24.100] _eventReaders is not empty after all the projections have been killed
could it be related to the problem ?
I’ve tested it with the v3.0.0rc2 and it works properly.
Thanks for the report we will take a look what the issue may be. Dev may also not be stable at this point. @yuriy do you have any known issues in dev now?
Cheers,
Greg