Permalink
Commits on Jul 22, 2011
  1. Increased ebm UPSTREAM_TIMEOUT to 12 hours

    Change-Id: I43137c52b71bc3b4ccdfcebc9ed39b6856d95812
    steveyen committed Jul 22, 2011
Commits on Jun 6, 2011
  1. Always use CHECKPOINT flag during TAP_CONNECT

    This is accomplised by always passing empty list [] instead of
    undefined for the vbucketId-to-checkpointId list.
    
    Change-Id: I3b5c80a3db6f6ec1848bfa5b116b5e219196c195
    Reviewed-on: http://review.couchbase.org/6798
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    steveyen committed with alk Jun 6, 2011
  2. TAP_CONNECT using checkpointId's of 0 rather than silence.

    Before this change, ebucketmigrator would TAP_CONNECT with a
    vbucketId-to-checkpointId pairs list whose length might be less than
    the vbucketId's list.
    
    Instead, this change has the ebucketmigrator TAP_CONNECT with 0's for
    checkpointId's for those vbucket's where it cannot determine the right
    last closed checkpointId for a vbucket.
    
    Change-Id: I76c94ba9644e81c3a39705c40fd6ab09177bfc77
    Reviewed-on: http://review.couchbase.org/6796
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    steveyen committed with alk Jun 6, 2011
Commits on Jun 4, 2011
  1. batch replication changes during rebalance

    So that we avoid too frequent re-connects to named tap streams and
    re-starts of backfill.
    
    The approach is to record replication changes and instead of applying
    them immediately, apply them only if a) there's some period of
    stability b) we need to apply at least one of them. Later is necessary
    in case where we need to stop replication to new master before doing
    takover into it. It is the only case where we need to sync data-safety
    wise because a) we don't reorder anything (apply all or delay all) b)
    only vbucket transfers can trigger bad thing (2 incoming streams for
    one vbucket)
    
    And if rebalance aborted (and we lose batched changes) then it's
    janitor's work to setup replication.
    
    Change-Id: I9866f22fa8d92a44b92b64bbd5f1632ec351c7a6
    Reviewed-on: http://review.couchbase.org/6769
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk Jun 4, 2011
  2. don't try to replicate from not-yet ready vbuckets

    Replicating from vbuckets being backfilled is a grey area and we don't
    want problems. So we detect this situation and do nothing for those
    vbuckets, but schedule retrying later.
    
    Change-Id: Iaf7797868c5380038bf67bc00b148d7dc6fd41ee
    Reviewed-on: http://review.couchbase.org/6758
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk Jun 3, 2011
Commits on May 31, 2011
  1. fixed swap percentage usage presentation

    Change-Id: If114f5445b4f131f2acaf9bf45e5cfed0d066219
    Reviewed-on: http://review.couchbase.org/6667
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 31, 2011
  2. removed tap queues fill rate from visible stats

    Because this stats are broken
    
    Change-Id: Ida7e2ac6b993a9ad50a3d1914de3ad97c0948b4f
    Reviewed-on: http://review.couchbase.org/6666
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 31, 2011
Commits on May 27, 2011
  1. implemented new computation of failover safeness level

    We still expose it as replication completion level as a number from 0
    to 1. Where 0 means no replication, and 1 means up-to-date w.r.t
    replication.
    
    Change-Id: Ib4a55e1bc2fb707a6f4241106966cd37cf481eb7
    Reviewed-on: http://review.membase.org/6580
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk May 27, 2011
  2. ensure minimal size of arcs on bucket readiness pie chart

    Based of work of Benjamin Young.
    
    We increase sizes of too small args at the 'expense' of largest
    arc. With only 3 arcs max it seems acceptable approach, even though
    it's not entirely cleanest.
    
    Change-Id: I8fb19175c0557e7d6f1583c6edaa9752365a7676
    Reviewed-on: http://review.membase.org/6611
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk May 27, 2011
  3. export force_beat from ns_heart

    So that anybody (like failover warning thing) can force state
    broadcast
    
    Change-Id: Id76638110930d1e6c75722cb1dd9b899e10050be
    Reviewed-on: http://review.membase.org/6578
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk May 27, 2011
  4. validate maximal replicaNumber

    Values more then 3 are not supported
    
    Change-Id: Idd783f7aa207cf9c7316cffa3ca1c6d639edf778
    Reviewed-on: http://review.membase.org/6584
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 26, 2011
  5. send fresh replication status with each hearbeat

    Previously it was part of expensive checks that we do less often.
    
    Change-Id: I32c60bc638226f497f43af02588a595e5b3d41c8
    Reviewed-on: http://review.membase.org/6579
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk May 27, 2011
  6. gather total_backlog_size stat of tap connections

    This stat will be used for failover safeness warnings
    
    Change-Id: I0f0ac241a8c0bd32085bc1a1ca15e5d55ec0657e
    Reviewed-on: http://review.membase.org/6577
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 27, 2011
  7. moved replication_status computation to ns_rebalancer

    Change-Id: I480660d8e62fd4c584c5808883dc3ba7b3ad8c43
    Reviewed-on: http://review.membase.org/6576
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk May 24, 2011
  8. fixed failover button width on non-mozilla browsers

    That float:right was completely useless.
    
    Change-Id: I482e9eab5b024fe46525a0c91d1ff5a77830c87c
    Reviewed-on: http://review.membase.org/6575
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 27, 2011
  9. CSS styles for warmup/unhealthy states on Fail Over buttons

    Change-Id: Ic9d424d1b0e13821a9ed88639872b0ea1f8ac2d2
    [alk@tut.by: changed class names slightly]
    Signed-off-by: Aliaksey Kandratsenka <alk@tut.by>
    Reviewed-on: http://review.membase.org/6534
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    BigBlueHat committed with alk May 25, 2011
  10. Rename sendStats config setting to send_stats.

    For a more consistent code base, I renamed the settings->stats->sendStats
    config key to send_stats. I think we should follow the rule:
    Within the Erlang world, use underscores as separator, for the
    outside world (JSON and POST parameters) use CamelCase.
    
    Change-Id: Ifecb0351ef20af51a12e2bd1e88c87aeed586cfd
    Reviewed-on: http://review.membase.org/6435
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    vmx committed with alk May 21, 2011
  11. un-disable vb_pending_ops_create

    Change-Id: I980bd9a5297b922fb8169495a52ec5fe0df9340a
    Reviewed-on: http://review.membase.org/6544
    Reviewed-by: Bin Cui <bin.cui@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 26, 2011
  12. fixed couple typos in stats descriptions (courtesy flyspell)

    Change-Id: Ibe5e2930541c25f1bd7ed4f0e6c5034994e937ad
    Reviewed-on: http://review.membase.org/6543
    Reviewed-by: Bin Cui <bin.cui@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 26, 2011
  13. stop fooling ourselves about net_ticktime of 10 seconds

    It was 10 seconds only on dev-mode clusters and (it appears)
    windows. Lowering it now seems a bit scary as most of our testing was
    for 60 seconds net_ticktime.
    
    So I'm fixing it (making option effective), but changing default to
    erlang's.
    
    Change-Id: Ia57713b54205e5b37052b1ba74fc08736867c099
    Reviewed-on: http://review.membase.org/6542
    Reviewed-by: Bin Cui <bin.cui@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 26, 2011
  14. fixed minor typo in stats title

    Change-Id: Ibd7cd73f0ccce8b2bdeca15e7e54302176263923
    Reviewed-on: http://review.membase.org/6541
    Reviewed-by: Bin Cui <bin.cui@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 26, 2011
  15. replaced total items column with replica items

    As requested in stats cleanup follow-up email.
    
    Change-Id: I75ede7adce5763b1b2ee44ecc4eb444b15973540
    Reviewed-on: http://review.membase.org/6540
    Reviewed-by: Bin Cui <bin.cui@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 26, 2011
  16. handle errors from failover and other clustering POSTs. MB-3767

    Change-Id: I636b22bd18f9b648fe4e00be859927c7f5b8ec4d
    Reviewed-on: http://review.membase.org/6525
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 25, 2011
  17. disable link to specific stats page for system stats

    Because it's not supported yet within REST. See MB-3726
    
    Change-Id: I90a31e7910f48660d25b02124e3edc35ac9bdfd2
    Reviewed-on: http://review.membase.org/6524
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 25, 2011
  18. strip hostname port in specific stats 'title's

    Change-Id: I42bc1eb0f940f5cb20f495e01317292440d97085
    Reviewed-on: http://review.membase.org/6523
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 25, 2011
  19. simplified and fixed hostname port cutting logic

    New logic is to strip port _only_ if all hostnames have default port
    (8091).
    
    So that if you have only single host with non-default port it's not
    hidden from you. This was especially problematic with single active
    host with port 9000 and single pending-add host with port 9001, in
    which case old logic was stripping port confusing you.
    
    Change-Id: I09f70703370f8020f61ea4d608d1758421544668
    Reviewed-on: http://review.membase.org/6522
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 25, 2011
  20. added connections stat to server stats

    Change-Id: I87b8c90cb797a4cab67c1fde5342cc0174ad6a57
    Reviewed-on: http://review.membase.org/6521
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 25, 2011
  21. implemented disk_write_queue stat

    Change-Id: I495ef1a1eaa639d35558f17b1bd29f3cbf9dc6ef
    Reviewed-on: http://review.membase.org/6520
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 25, 2011
  22. get rid of null sample values in collector

    This fixes badarith in menelaus_stats:basic_stats.
    
    We're converting them to 0 anyways so it makes little sense.
    
    The only place in stats that still produces null values is
    samples_to_proplists which is final data transformation before sending
    it to client. And 'null' is used to mark missing values, which may
    happen after membase upgrade.
    
    Change-Id: Ib9e1e94179fe6fc718e42f264b2bbb7ee880fe00
    Reviewed-on: http://review.membase.org/6514
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 24, 2011
  23. unbreak handling of improper lists in diag_handler

    With recent change we now have improper lists for many values, and
    this breaks assumptions of
    diag_handler:diag_filter_out_config_password_list. This fix makes
    obvious correction.
    
    Change-Id: Iaf5032012e0f250c3cee7522656ac1813626d7d5
    Reviewed-on: http://review.membase.org/6513
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 24, 2011
  24. Separate static config data dir from bin dir.

    In at least the OS X install, the data and bin locations are
    different.
    
    Change-Id: I2f21248324eed6df7d037ec2b6842c9014bdc774
    Reviewed-on: http://review.membase.org/6601
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    dustin committed with alk May 27, 2011
  25. increase ebucketmigrator upstream timeout to 5 minutes

    Because sometimes when tap producer is very busy with fetching things
    from disk, 30 seconds might be not enough.
    
    Change-Id: I1d51233ec5bd6ba8d5a8e926afa422baf08c301f
    Reviewed-on: http://review.membase.org/6573
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk May 27, 2011
Commits on May 26, 2011
  1. Fix race condition inside rebalance

    This can also be fixed by moving vbucket generation to after
    create_bucket, however that would require a large restructuring
    of bucket_sup
    
    Change-Id: Ifd3b9d58924bb5693a86e28381c9e42ecf7e4b4b
    Reviewed-on: http://review.membase.org/6538
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Dale Harvey committed with alk May 25, 2011
  2. access previous vbucket state

    Change-Id: I253ffd98efa159bf73099c7ae4da0dcd5b21da97
    Reviewed-on: http://review.membase.org/6530
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Dale Harvey committed with alk May 25, 2011
  3. use default data layout

    Change-Id: Ibfa338666246c4cfeecc537c6afceb214bfb0bf6
    Reviewed-on: http://review.membase.org/6529
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Dale Harvey committed with alk May 25, 2011