DR102 too much data written uncommitted

[Michael Henson]

We've seen a few instances of the following warning message followed by the assertion below. This has only happened on one of our machines which is a hidden and unloaded member of a mongo replicaset running 2.0.7. I have not seen the error on our other members. Is there anything we can do to help track down the cause of this?

[rsSync] warning: DR102 too much data written uncommitted 314.577MB

0x6fdd79 0x6f249a 0x7b3a9c 0x7bb741 0x7bc0c1 0x7b3546 0x7b3b6b 0x7ec5c2 0x791966 0x78c73c 0x78d1be 0x78d

292 0x78d707 0x7f826e223b70 0x7f826e94c9ca 0x7f826d67ecdd 

/usr/bin/mongod(_ZN5mongo3dur9CommitJob4noteEPvi+0x279) [0x6fdd79]

/usr/bin/mongod(_ZN5mongo3dur11DurableImpl10writingPtrEPvj+0xa) [0x6f249a]

/usr/bin/mongod(_ZN5mongo16NamespaceDetails13addDeletedRecEPNS_13DeletedRecordENS_7DiskLocE+0x21c) [0x7b3a9c]

/usr/bin/mongod(_ZN5mongo16NamespaceDetails7compactEv+0x241) [0x7bb741]

/usr/bin/mongod(_ZN5mongo16NamespaceDetails11cappedAllocEPKci+0x271) [0x7bc0c1]

/usr/bin/mongod(_ZN5mongo16NamespaceDetails6_allocEPKci+0x16) [0x7b3546]

/usr/bin/mongod(_ZN5mongo16NamespaceDetails5allocEPKciRNS_7DiskLocE+0x3b) [0x7b3b6b]

/usr/bin/mongod(_ZN5mongo11DataFileMgr17fast_oplog_insertEPNS_16NamespaceDetailsEPKci+0x162) [0x7ec5c2]

/usr/bin/mongod(_ZN5mongo11_logOpObjRSERKNS_7BSONObjE+0x236) [0x791966]

/usr/bin/mongod(_ZN5mongo11ReplSetImpl8syncTailEv+0xa9c) [0x78c73c]

/usr/bin/mongod(_ZN5mongo11ReplSetImpl11_syncThreadEv+0x7e) [0x78d1be]

/usr/bin/mongod(_ZN5mongo11ReplSetImpl10syncThreadEv+0xa2) [0x78d292]

/usr/bin/mongod(_ZN5mongo15startSyncThreadEv+0x97) [0x78d707]

/usr/lib/libboost_thread.so.1.40.0(thread_proxy+0x60) [0x7f826e223b70]

/lib/libpthread.so.0(+0x69ca) [0x7f826e94c9ca]

/lib/libc.so.6(clone+0x6d) [0x7f826d67ecdd]

[Gianfranco]

This seems to a problem which will be fixed in 2.2.1:

https://jira.mongodb.org/browse/SERVER-6816

This happened only once?

[Michael Henson]

I've seen this happen on two different days. On the first day it happened 3 times in rapid succession. On the second day it happened just once. Is there any chance that there could be data loss as a result of this?

[Gianfranco]

Sorry I missed that from your initial message.

There is a possibility that some extra data won't be written to disk.

The warning on the log is not fatal and it can be prevented improving disk io or throttling writes.