Skip to content
Commits on Aug 19, 2012
  1. MB-6216: validate all replicators connections each janitor run

    Aliaksey Kandratsenka committed with Peter Wansch Aug 19, 2012
    So that rebalance cannot fail due to failure to change vbucket on
    stale connection.
    
    Change-Id: I49718edf22bc02eee42abbe8b3fadae4493782cc
    Reviewed-on: http://review.couchbase.org/19848
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
Commits on Aug 18, 2012
  1. MB-6284: don't touch index pausing API api if we don't need it

    Aliaksey Kandratsenka committed with Peter Wansch Aug 17, 2012
    There's some evidence that NOP mark_partitions_indexable call takes
    few seconds.
    
    Also given that if index_aware_rebalance_disabled is true, we don't
    pause indexes, it makes sense to avoid calling those NOP calls too.
    
    Change-Id: Icdffd7f73c77526c14a21a0c942f7345d4731c35
    Reviewed-on: http://review.couchbase.org/19811
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  2. MB-6284: bumped set_vbucket_state_timeout to infinity

    Aliaksey Kandratsenka committed with Peter Wansch Aug 17, 2012
    Because we're using this call as part of rebalance which is explicitly
    started and explicitly stopped. There's no need to have any timeouts.
    
    Especially given that as part of this call we're doing state changes
    for potentially lots and lots of indexes.
    
    Change-Id: Iddb7194b2b3718e0996c987c4f99dd4ee5fc8dd7
    Reviewed-on: http://review.couchbase.org/19812
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
Commits on Aug 17, 2012
  1. bp: MB-5307: log usable vbuckets set

    Aliaksey Kandratsenka committed with Peter Wansch Aug 16, 2012
    So that we can diagnose cases like MB-6049 better.
    
    Change-Id: Id151229c974fa5c6f05d59750c8d051a08c81b0d
    Orig-Reviewed-on: http://review.couchbase.org/19734
    Reviewed-on: http://review.couchbase.org/19783
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  2. bp: MB-5307: fixed periodic logging of stats during rebalance

    Aliaksey Kandratsenka committed with Peter Wansch Aug 16, 2012
    Change-Id: Iaf1db5b3bc01fc2f87f9ff0efa7a3fb827a21cdf
    Orig-Reviewed-on: http://review.couchbase.org/19733
    Reviewed-on: http://review.couchbase.org/19782
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  3. bp: MB-5307: mark prerelease version in config

    Aliaksey Kandratsenka committed with Peter Wansch Aug 16, 2012
    So that we have chance of being able to online/offline upgrade from
    this config/environment in some later version.
    
    Change-Id: I73b75253197908785ee45842ed9938fe96386039
    Orig-Reviewed-on: http://review.couchbase.org/19714
    Reviewed-on: http://review.couchbase.org/19763
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  4. bp: MB-6256: enable basic auth on capi port

    Aliaksey Kandratsenka committed with Farshid Ghods Aug 16, 2012
    Change-Id: I846ce3468b39b5b82964eaafe9201be552f85990
    Orig-Reviewed-on: http://review.couchbase.org/19752
    Reviewed-on: http://review.couchbase.org/19764
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Farshid Ghods <farshid@couchbase.com>
    Tested-by: Farshid Ghods <farshid@couchbase.com>
  5. bp: MB-5307: fix set_view_update_daemon caused by recent changes

    Aliaksey Kandratsenka committed with Farshid Ghods Aug 16, 2012
    When we've switched to use of ddocs stored inside
    capi_ddoc_replication_srv this was broken.
    
    Problem happens because ddocs from interactive_updates are with binary
    bodies and ddocs we need are with ejson bodies.
    
    Change-Id: I5fb8ba303aa04d410a1610154e6478cdfe9902de
    Orig-Reviewed-on: http://review.couchbase.org/19713
    Reviewed-on: http://review.couchbase.org/19762
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Farshid Ghods <farshid@couchbase.com>
    Tested-by: Farshid Ghods <farshid@couchbase.com>
Commits on Aug 15, 2012
  1. @alk

    MB-6160: apply index config changes on ddoc replication process

    Aliaksey Kandratsenka committed with alk Aug 14, 2012
    So that we can stop worrying if some ddoc becomes deleted concurrently
    with us trying to do something with it.
    
    Notable side effects are:
    
    * cb_generic_replication_srv keeps full #doc-s in it's state
    
    * capi_set_view_manager does not keep track of all ddocs anymore,
      relying on capi_ddoc_replication_srv
    
    Change-Id: I1a02af2968d9321f9d21824493a6bb9f0710451b
    Reviewed-on: http://review.couchbase.org/19646
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  2. @alk

    MB-5308: display correct servers count for bucket

    Aliaksey Kandratsenka committed with alk Aug 14, 2012
    I.e. using it's real used servers list rather than all known servers.
    
    Change-Id: Ic98b9bc5e25284cf83923251e9c783ddc51c4537
    Reviewed-on: http://review.couchbase.org/19645
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  3. @alk

    MB-6042: avoid retries and excessive work in vbucket_map_mirror

    Aliaksey Kandratsenka committed with alk Aug 14, 2012
    When lots of clients spot lack of cached nodes->vbuckets dictionary
    they will all call into vbucket_map_mirror worker. Previous version
    would recompute map for each such call. Clearly, not perfect approach,
    given that worker has full access to up-to-date cached version.
    
    So new code is just reusing already cached version if it's there at
    the time worker call is processed. This changes top-level logic a bit,
    as call_compute_map now returns dict. So there's no need to retry.
    
    As a side effect we're now correctly handling a case when bucket
    disappeared right before call into node_vbuckets_dict which is I think
    what actually happened in MB-6042.
    
    Change-Id: I151791a3898995b3096bba2d55fe0205945629d4
    Reviewed-on: http://review.couchbase.org/19618
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  4. @alk

    MB-5307: correctly handle in-shutdown-buckets in diag_handler

    Aliaksey Kandratsenka committed with alk Aug 14, 2012
    Change-Id: Id53656914606de30f365a9cda1666c3a31699677
    Reviewed-on: http://review.couchbase.org/19613
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  5. @alk

    MB-5307: don't crash per-node diag on tap stats exception

    Aliaksey Kandratsenka committed with alk Aug 14, 2012
    Because there's tons of other very useful information we need.
    
    Change-Id: I826ef012b9d5ca6d6af43416faf0f9233bde5abe
    Reviewed-on: http://review.couchbase.org/19612
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
Commits on Aug 14, 2012
  1. @alk

    MB-6206: fixed node membership check race when starting rebalance

    Aliaksey Kandratsenka committed with alk Aug 14, 2012
    All checks are now done on orchestrator, serializing things
    w.r.t. node state changes.
    
    Also it can be seen that previosuly no_active_nodes_left check was not
    taking into account failed over node which made possible MB-6206,
    i.e. rebalancing out all the nodes.
    
    Change-Id: I8933407aafaf28914c4615e45a65858255c95328
    Reviewed-on: http://review.couchbase.org/19614
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
Commits on Aug 13, 2012
  1. @alk

    MB-5308: correctly present filter text as well

    Aliaksey Kandratsenka committed with alk Aug 13, 2012
    By not url-un-espaping it either. So that things like + and & stay
    escaped to prevent confusion.
    
    Change-Id: If21bff3f19d6f665964b27013f57ab08a41568ec
    Reviewed-on: http://review.couchbase.org/19574
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  2. @pavel-blagodov @alk

    MB-6166: removed decodeURIcomponent function

    pavel-blagodov committed with alk Aug 13, 2012
    jQuery deparam contains decodeURIcomponent,
    that's why I removed this function.
    
    Change-Id: I5c199bcc7f05aba21ed6e37fe683508c154cc06b
    Reviewed-on: http://review.couchbase.org/19555
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  3. @alk

    MB-6115: fix packing order issue in memcached binary protocol

    Junyi Xie committed with alk Aug 10, 2012
    Today during XDCR, CAPI creates a packet using binary protocol and
    sends it to ep_engine. The order CAPI packs the item metadata in the
    extension part is expiration time followed by flags, which is however
    inconsistent with ep_engine protocol since it puts flags ahead of
    the expiration time. This change-set will fix the order issue.
    
    Change-Id: Ia7e76ae34d9812d73779c428c99a2d810366277c
    Reviewed-on: http://review.couchbase.org/19508
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  4. @aartamonau @alk

    MB-5307 Remove unneeded init.sql.

    aartamonau committed with alk Aug 13, 2012
    Change-Id: I235cc240b76185b864dcb8f858d1081d363d5749
    Reviewed-on: http://review.couchbase.org/19560
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  5. @aartamonau @alk

    MB-6032 Upgrade memcached config on 1.8.1 to 2.0 offline upgrade.

    aartamonau committed with alk Aug 13, 2012
    Particularly mccouch port and logger parameters are added to the
    config and old engines are replaced with new ones.
    
    Change-Id: I8b61235c92f0208585f7b213b1b6046abc9e3e8a
    Reviewed-on: http://review.couchbase.org/19559
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  6. @aartamonau @alk

    MB-5307 Name :one_for_one uniformly in server hierarchy.

    aartamonau committed with alk Aug 13, 2012
    Change-Id: I17c6810be35d11c2d37072802127f50ac86f5b91
    Reviewed-on: http://review.couchbase.org/19558
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  7. @aartamonau @alk

    MB-5307 Fix 1.8.1 -> 2.0 upgrade function name.

    aartamonau committed with alk Aug 13, 2012
    Change-Id: I34900ee585927c03dfa635ca1240c8c4b9d508c8
    Reviewed-on: http://review.couchbase.org/19557
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  8. @aartamonau

    MB-5307: added milliseconds to logger timestamps

    Aliaksey Kandratsenka committed with aartamonau Jul 30, 2012
    Because it's very convenient to have milliseconds in log entries.
    
    This commit also makes log message lines narrower so that more stuff
    fits into screen.
    
    Change-Id: I30383234f83e90fd941374881e529eac786410f8
    Reviewed-on: http://review.couchbase.org/19027
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  9. @aartamonau

    MB-6201: delete bucket synchronously and return 500 on time out

    Aliaksey Kandratsenka committed with aartamonau Aug 11, 2012
    Change-Id: Idee0de1a82811d353ba63a8ba4b4b3e84b14a209
    Reviewed-on: http://review.couchbase.org/19520
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  10. @aartamonau

    MB-6200: correctly check if bucket is still being deleted

    Aliaksey Kandratsenka committed with aartamonau Aug 11, 2012
    Change-Id: I6ba393a48f8cdeb8b7b0296cde7bfe703b1b84a3
    Reviewed-on: http://review.couchbase.org/19519
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  11. @aartamonau

    MB-5307: crash ebucketmigrator on seeing any exit signal

    Aliaksey Kandratsenka committed with aartamonau Aug 10, 2012
    As can be seen in logs of MB-6058 if 'vbucket filter change
    transaction' process that's linked to ebucketmigrator fails, we need
    to fail bucketmigrator too. Which did not happen previously.
    
    Change-Id: Ib657edfc14cf51af33bb9c88ca7b0831d00b21c1
    Reviewed-on: http://review.couchbase.org/19518
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  12. @aartamonau

    MB-5859: allow more time for pre-rebalance janitor run

    Aliaksey Kandratsenka committed with aartamonau Aug 10, 2012
    Our past expectation was to wait for memcached on all nodes and that
    following janitor run does not need to wait. But in some cases there
    is some variability and 1 second timeout is not enough. Thus we raise
    it.
    
    Change-Id: Icbaedfd19d46279de496e8f1928c2b1a083fec5d
    Reviewed-on: http://review.couchbase.org/19517
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  13. @aartamonau

    MB-5307: log message before going into vbucket dbs deletion

    Aliaksey Kandratsenka committed with aartamonau Aug 10, 2012
    So that we have better understanding of certain problems.
    
    Change-Id: I255b174032da17c705dcfdf5bbd37f6ecb79a4b3
    Reviewed-on: http://review.couchbase.org/19516
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  14. @aartamonau

    MB-5307: killed remnant of multiple ns_memcached's per bucket

    Aliaksey Kandratsenka committed with aartamonau Aug 10, 2012
    Change-Id: I299aeaf1ac24d3c754c761847750b1a7a02202b7
    Reviewed-on: http://review.couchbase.org/19515
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
Commits on Aug 12, 2012
  1. @alk

    MB-6167: fixed saving map/reduce functions into wrong views

    Aliaksey Kandratsenka committed with alk Aug 11, 2012
    This was caused by idiotic behavior of lazy Cells that I introduced
    some time ago. See test case that's now fails.
    
    Basicaly I wanted re-subscribtion to detached cells to not recompute
    formulas and not re-fetch values (if formula computes to future). And
    that's causing current code to keep old value when cell is
    detached. And that in turn caused #getValue to just use old value
    instead of triggering cell recomputation.
    
    Because fixing it either requires bigger changes or change of behavior
    it's a bit too scary for now. Thus I added failing test and changed
    affected cells to eager type.
    
    Change-Id: I9c181d89e62df2dbd8ed9fe692b1a6cbbf51b574
    Reviewed-on: http://review.couchbase.org/19514
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
Commits on Aug 11, 2012
  1. Merge remote-tracking branch 'origin/branch-18'

    Aliaksey Kandratsenka committed Aug 11, 2012
    * origin/branch-18:
      MB-5895: consider upgrading vb map history at rebalance start
    
    Change-Id: I883b51f85a45cdd8c9e0dce7738207507b91b8be
    NOTE: diff is empty as 2.0 already fixed MB-5895 in different way
  2. @alk

    Merge "MB-100: Merge commit 'f2930c1e14310a95fc4f7fd14275f53685babdf0'"

    alk committed with Gerrit Code Review Aug 11, 2012
  3. @alk

    MB-6179: don't include master vbucket into all_docs

    Aliaksey Kandratsenka committed with alk Aug 10, 2012
    Those are 2 different namespaces since some time ago.
    
    Change-Id: I3dcaa97ad43a48103a7db58e183ba3f8daa43460
    Reviewed-on: http://review.couchbase.org/19474
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  4. @alk

    MB-5307: implemented semi-official API for resetting node's state

    Aliaksey Kandratsenka committed with alk Aug 10, 2012
    Posting otpNode=zzzForce to /controller/ejectNode is going to cause
    node to leave cluster, completely resetting it's state as if it was
    rebalanced out.
    
    Call will return _after_ eject completes and ejected node is ready to
    process further requests.
    
    Change-Id: Icac3c7466d461d09fba90c6a27884813b90e415f
    Reviewed-on: http://review.couchbase.org/19473
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
  5. @alk

    CBD-178: implemented index-aware vbucket moving

    Aliaksey Kandratsenka committed with alk Aug 7, 2012
    In order to have 'consistent views' during rebalance we need to make
    sure indexes for vbucket being moved are at least as up to date on new
    master as on old master.
    
    The following is performed to achieve that:
    
    * on replicas and future master we're waiting to persist at least same
      docs as were persisted on old master
    
    * indexing on old master is paused for current vbucket to ensure index
      does not 'move forward' w.r.t. this vbucket
    
    * we're waiting again while future master has persisted _and_ indexed
      (in passive state) at least same docs as were persisted on old
      master after pausing indexing.
    
    * we perform vbucket takeover, switch vbucket into active state on new
      master and cleanup state on old master, then update vbucket
      map. From this point, queries will use vbucket's index on new
      master.
    
    As can be seen, this implements 'monotonicity' of index from the point
    of queries reading index.
    
    New behavior is disable-able by ns_config variables:
    
    * index_aware_rebalance_disabled
    
    * index_pausing_disabled (just disabling pausing)
    
    As of this commit both are disabled to enable fuller testing before
    enabling them.
    
    Change-Id: I732bb225673e8e8969747fc83fdc557b124b1a6c
    Reviewed-on: http://review.couchbase.org/19347
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
  6. @alk

    MB-5307: improved capi_set_view_manager typespecs

    Aliaksey Kandratsenka committed with alk Aug 7, 2012
    Change-Id: I5a530dbb04a79a1bdd9674727b1c11ae24293c96
    Reviewed-on: http://review.couchbase.org/19346
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
Something went wrong with that request. Please try again.