[MB-4717] flush_all doesn't work on the latest 2.0 builds Created: 30/Jan/12  Updated: 19/Jul/14  Resolved: 03/Aug/12

Status: Closed
Project: Couchbase Server
Component/s: couchbase-bucket
Affects Version/s: 2.0-developer-preview-4
Fix Version/s: 2.0-beta
Security Level: Public

Type: Bug Priority: Major
Reporter: Farshid Ghods (Inactive) Assignee: Dipti Borkar
Resolution: Duplicate Votes: 0
Labels: 2.0-dev-preview-4-release-notes
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
is duplicated by MB-5046 inconsistent_state error on productio... Closed

some of the failures we saw recently in couch were caused by flush _alll

will add more info later

Comment by Farshid Ghods (Inactive) [ 13/Mar/12 ]
flush_all is not handled correctly in 2.0 dp4 builds if there are views defined in the node/cluster or if flush is invoked when view or database compaction is running.

users should use bucket_deletion for deleting and recreating the bucket.
Comment by Filipe Manana (Inactive) [ 02/May/12 ]
Farshid, from what I understand the flush command is no longer supported for DP5.
If true, shall this issue be closed?
Comment by Farshid Ghods (Inactive) [ 03/Jun/12 ]
flush_all is disabled by default but the user can enable it and use it for testing
Comment by Matt Ingenthron [ 03/Jun/12 ]
For 2.0, the plan is to make the REST flush work correctly (though off by default) and leave this flush disabled. I don't know of an issue tracking this currently.
Comment by Matt Ingenthron [ 03/Jun/12 ]
Farshid: could you clarify the description here? Are you talking about 'failures in couch' or users seeing their data lost due to inadvertently issuing a flush_all command?
Comment by Farshid Ghods (Inactive) [ 03/Jun/12 ]
this is failure in couch and compaction daemons after calling flush_all. ( so the user won't be able to use the bucket after flush )
Comment by Matt Ingenthron [ 07/Jun/12 ]
MB-5170 is tracking the implementation of a REST level flush feature.
Comment by Peter Wansch (Inactive) [ 28/Jun/12 ]
Can you update the affects and fixed version and if you know this is a current or next sprint item, the sprint status as well? Thank you.
Comment by Filipe Manana (Inactive) [ 28/Jun/12 ]
No idea what this is about neither how the view engine could solve anything here.
This seems like some component (mccouch?) is not doing the necessary view engine interaction.
Comment by Farshid Ghods (Inactive) [ 24/Jul/12 ]
assigning this to Dipti.

do we plan to support flush in 2.0 ?
Comment by Matt Ingenthron [ 24/Jul/12 ]
To my knowledge, plan is the same as described in my comment on June 3. In more detail, plan is to:
1) disable flush_all by default, with a method to enable it with known drawbacks on it not being synchronous (but not broken) until...
2) RESTful flush is fixed to be correct to its interface and at least reasonable latency (and still have it off by default) and then
3) Migrate all tests to use the now reliable RESTful flush
4) Take away the method to disable flush_all on a per-node basis
Comment by Trond Norbye [ 30/Jul/12 ]
this isn't a bucket_engine thing (works as expected for memcached buckets), but a problem with the couchdb buckets
Comment by Dipti Borkar [ 03/Aug/12 ]
Generated at Wed Nov 26 17:34:50 CST 2014 using JIRA 5.2.4#845-sha1:c9f4cc41abe72fb236945343a1f485c2c844dac9.