Details
-
Bug
-
Resolution: Fixed
-
Blocker
-
None
-
3.0
-
None
-
Security Level: Public
-
None
-
Untriaged
Description
The view engine uses couch_db:get_purge_seq(Db) to get the purge sequence number and checks if the indexed sequence number is greater than purge sequence number before proceeding to start reading sequences of changes happened after last index update (This will not be applicable once indexer start reading through UPR). It is noticed that the couch_db:get_purge_seq(Db) returns some out of int range type of numbers and it causing view engine to do full index build from scratch.
Eg. seq no retreived is 34625945600
Logs:
[couchdb:info,2014-02-06T14:02:36.540,n_0@192.168.1.120:<0.5565.0>:couch_log:info:39]Updater reading changes from active partitions to update main set view group `_design/test_view-faaefe5` from set `default`¬
[couchdb:info,2014-02-06T14:02:36.540,n_0@192.168.1.120:<0.5565.0>:couch_log:info:39]Updater main set view group `_design/test_view-faaefe5` from set `default`, exiting with purge reason because partition 0 has purge seq 34625945600 and last indexed seq is 638¬
[couchdb:info,2014-02-06T14:02:36.545,n_0@192.168.1.120:<0.4026.0>:couch_log:info:39]192.168.1.120 - - GET /default/_design/test_view-faaefe5/_view/test_view-faaefe5?startkey=%5B2008%2C7%2Cnull%5D&connection_timeout=60000 200¬
[couchdb:info,2014-02-06T14:02:36.561,n_0@192.168.1.120:<0.1310.0>:couch_log:info:39]Set view `default`, main (prod) group `_design/test_view-faaefe5`, group reset because updater detected missed document deletes (purge)
We did not make any changes in CouchDB erlang client. So, I doubt if some header structure refactoring has taken place for Couchstore wrt UPR.
Attachments
Issue Links
- blocks
-
MB-10133 View query tests fail with error on {u'reason': u'{badmatch,eof}', u'from': u'local'}
- Closed