I am trying to repair a single CF using nodetool. It seems like the request
to limit the repair to one CF is not being respected. Here is my current
situation:
- Run "nodetool repair KEYSPACE CF_A" on node 3
- Validation compaction runs on nodes 2,3,4 for CF_A only
(expected)
- Node 3 streams SSTables from CF_A only to nodes 2 and 4
(expected)
- Nodes 2 and 4 stream SSTables from ALL column families in the
keyspace to node 2 (VERY unexpected)
- Node 2 runs out of disk space before SSTable rebuild for all cfs
can complete.
Presumably this is a bug (?). This is the first time in quite awhile that I
have run a repair (I don't perform deletes, just use expiring columns). I
have included pertinent log entries below:
Log entry from node 3 after validation compaction (I would say expected):
INFO [AntiEntropyStage:1] 2011-01-24 13:36:06,235 AntiEntropyService.java
(line 220) Queueing comparison #<Differencer #<TreeRequest
manual-repair-ef684731-cb9b-408c-acaa-094234e58979, /192.168.4.16,
(kikmetrics,UserEventsByUser)>>
INFO [AntiEntropyStage:1] 2011-01-24 13:36:06,624 AntiEntropyService.java
(line 481) Endpoints /192.168.4.17 and /192.168.4.16 have 1 range(s) out of
sync for (kikmetrics,UserEventsByUser)
INFO [AntiEntropyStage:1] 2011-01-24 13:36:06,625 AntiEntropyService.java
(line 498) Performing streaming repair of 1 ranges for #<TreeRequest
manual-repair-ef684731-cb9b-408c-acaa-094234e58979, /192.168.4.16,
(kikmetrics,UserEventsByUser)>
INFO [AntiEntropyStage:1] 2011-01-24 13:36:06,655 StreamOut.java (line 173)
Stream context metadata
[/var/lib/cassandra/data/kikmetrics/UserEventsByUser-e-3-Data.db/(0,12858523
5)
progress=0/128585235 - 0%,
/var/lib/cassandra/data/kikmetrics/UserEventsByUser-e-2-Data.db/(0,34069035)
progress=0/34069035 - 0%,
/var/lib/cassandra/data/kikmetrics/UserEventsByUser-e-1-Data.db/(0,31673166)
progress=0/31673166 - 0%,
/var/lib/cassandra/data/kikmetrics/UserEventsByUser-e-4-Data.db/(0,25515348)
progress=0/25515348 - 0%], 4 sstables.
Log entry from node 2/4 after validation (unexpected, it is sending sstables
for all CFs, not just the one being repaired):
INFO [StreamStage:1] 2011-01-24 13:36:42,815 StreamOut.java (line 173)
Stream context metadata
[/var/lib/cassandra/data/kikmetrics/PacketEventsByEvent-e-1573-Data.db/(8501
3299,170188890)
progress=0/85175591 - 0%,
/var/lib/cassandra/data/kikmetrics/PacketEventsByEvent-e-1559-Data.db/(86608
8590,1569557337)
progress=0/703468747 - 0%,
/var/lib/cassandra/data/kikmetrics/PacketEventsByEvent-e-1519-Data.db/(57433
2782,1027659278)
....
progress=0/5158259489 - 0%], 77 sstables.
Log entry from node 3 after running out of disk space:
INFO [Thread-3462] 2011-01-24 16:00:40,658 StreamInSession.java (line 124)
Streaming of file
/var/lib/cassandra/data/kikmetrics/UserEventsByEvent-e-1313-Data.db/(0,28295
156514)
progress=15267921920/28295156514 - 53% from
org.apache.cassandra.streaming.StreamInSession@8df7e0c failed: requesting a
retry.
INFO [Thread-3423] 2011-01-24 16:00:40,658 StreamInSession.java (line 124)
Streaming of file
/var/lib/cassandra/data/kikmetrics/UserEventsByEvent-e-1348-Data.db/(2953062
0905,58066315307)
progress=18898300928/28535694402 - 66% from
org.apache.cassandra.streaming.StreamInSession@8df7e0c failed: requesting a
retry.
Dan Hendry
(403) 660-2297