#opendaylight-clustering: clustering_hackers

Meeting summary

1. bugs (moizer, 16:52:16)
   1. Mark was able to reproduce issue with bug 2653 using clustering in Helium SR2. Need to figure out how to provide a patch for it. (moizer, 16:53:43)
   2. Moiz still playing around with 3-Node cluster and trying to find the optimal back-pressure to apply and the optimal election timeout factor (moizer, 16:54:40)
   3. Tom wan’t to start a RaftActor as a Leader if there are no Followers. Will create a bug for it. (moizer, 16:55:16)
   
   
2. reviews (moizer, 16:55:32)
   1. Tom to merge Kamal’s changes today if possible as Kamal will be out for next 10 days (moizer, 16:55:59)
   2. Tom has a patch on batching operations in a transaction to reduce the number of messages that flow between TransactionProxy to ShardTransaction - added more eyes to it (moizer, 16:57:02)
   
   
3. Vamsi observations on Migration to akka 2.3.9 and clustered data store performance (moizer, 16:58:15)
   1. Had to reduce the list elements per transaction to 1,000 from 100,000 to make test go through (moizer, 16:59:13)
   2. Performance limited by I/O. All transactions were done on the config datastore (moizer, 16:59:46)
   3. 1 transaction with 100,000 operations takes in the order of 152s (moizer, 17:00:20)
   4. Need to understand if this is pure I/O related or something else is in play (moizer, 17:00:37)
   5. Tom thinks batching operations may help as it reduces the communication between Transaction Proxy and ShardTransaction and Serialization is more efficient (moizer, 17:01:53)
   6. Switching from akka 2.3.4 -> 2.3.9 appears to be safe. Vamsi to propose a patch for this upgrade. (moizer, 17:03:13)

Action items

1. (none)

People present (lines said)

1. moizer (18)
2. odl_meetbot (4)
3. tbachman (3)
4. ttkacik1 (2)