Permalink
Commits on Dec 29, 2011
  1. Kill upstream_sender process explicitly.

    This avoids leaking processes when ebucketmigrator_srv terminates with
    reason 'normal'.
    
    Change-Id: I7b36573ad4a206482e15167a0c76b6b284b9cead
    Reviewed-on: http://review.couchbase.org/11952
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    aartamonau committed with alk Dec 29, 2011
  2. Crash if message can't be delivered to upstream.

    Change-Id: Id7d2b8be5d3ada55c4094195ecdda3ab5f3ce67b
    Reviewed-on: http://review.couchbase.org/11951
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    aartamonau committed with alk Dec 28, 2011
Commits on Dec 27, 2011
  1. Always keep current node in the list of new peers in mb_master.

    Fix for MB-4592.
    
    Change-Id: If3e4c1011618f976d9932ff8b46ccc62856ab183
    Reviewed-on: http://review.couchbase.org/11907
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    aartamonau committed Dec 27, 2011
  2. don't eject failed over nodes twice

    This is unnecessary. May cause data loss on node being failed over (if
    some automated system tries doing anything on it right after it being
    failed over). And currently triggers memcached init/shutdown race (see
    memcached commit c8f5e90ff432d2f3b4db1991c403b8fa9824d284).
    
    Change-Id: Ib1558fc3a902ceeb64dae48d84b6e952542da99d
    Reviewed-on: http://review.couchbase.org/11905
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Aliaksey Kandratsenka committed with aartamonau Dec 27, 2011
  3. made janitor drop rebalance status iff really not running.MB-4579

    maybe_stop_replication_status was assuming if janitor is run,
    replication is not running. So we simply reset replication status.
    
    But we in fact do call janitor as part of rebalance of each bucket. So
    janitor was incorrectly resetting replication status to "not running".
    
    Fix is to check in orchestrator (which should be local) if it's really
    running rebalance or not.
    
    Change-Id: I890a5f8739bae7be1da562a51e9ae4d5d40bf028
    Reviewed-on: http://review.couchbase.org/11904
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Aliaksey Kandratsenka committed with aartamonau Dec 27, 2011
Commits on Dec 24, 2011
  1. reset replication_status when janitor sees it as running

    When node orchestrating rebalance goes down or entire cluster is
    rebooted during rebalance, we can get replication_status in config in
    state running.
    
    In order to get most reliable fix we're resetting status at the end of
    janitor run, if janitor sees all nodes as up. That minimizes chance of
    network split or any other transient problem causing rebalance_status
    reset.
    
    Change-Id: I02d56b8d9d4990442498e08a92db5fce49e5bf07
    Reviewed-on: http://review.couchbase.org/11870
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Aliaksey Kandratsenka committed with aartamonau Dec 23, 2011
Commits on Dec 23, 2011
  1. made ns_memcached log slow calls

    Change-Id: I7db7582629db3f684f165e09079624ba3a622a73
    Reviewed-on: http://review.couchbase.org/11869
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 23, 2011
  2. made ns_doctor: log ready -> down/warmup bucket state transitions

    Change-Id: I18cd098481408b61da63de92cc56c52104cfa469
    Reviewed-on: http://review.couchbase.org/11868
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 23, 2011
  3. increased ns_port_server log throttling limits

    So that we can see more ep-engine/memcached log messages.
    
    I'm also increasing ring buffer size of post-mortem logging to 100
    lines. That should still be cheap enough, but preserve more
    ultra-important diagnostics.
    
    Change-Id: If3706282e1da96535947eedce6909d5088f157f6
    Reviewed-on: http://review.couchbase.org/11867
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 23, 2011
  4. added tooltip text for plant my tree feature

    Change-Id: Ic22e0126a16cf5134cd4753e77c15b0fed5fcd83
    Reviewed-on: http://review.couchbase.org/11835
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 22, 2011
  5. send acks to upstream via separate process

    This fixes some rare cases where upstream memcached tries to send TAP
    data and does not receive ACKS/NACKS we're trying to send it. This
    leads to deadlock where we're waiting for ack to be sent (blocking on
    memcached consuming some of it's queued nacks/acks) and memcached waiting on
    us receiving TAP data.
    
    TAP ack design requires us to be able to queue up to tap_ack_interval
    nacks. So doing downstream -> upstream path asynchronously achieves
    that. Without, hopefully, introducing any really excessive
    queuing. Because of, seemingly, limited amount of queued acks/nacks
    required.
    
    NOTE: when we're shutting ebucketmigrator down, we're dropping queued
    to-upstream data on the floor. But that's what essentially happened
    before anyways.
    
    Change-Id: I1a518e55e3a539976f921dea1afc0c7ccff15f48
    Reviewed-on: http://review.couchbase.org/11859
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 23, 2011
Commits on Dec 22, 2011
  1. unbreak enter in second wizard step in IE8

    Change-Id: I43dabe0242faf41e620e5d34d845934a927dc3a7
    Reviewed-on: http://review.couchbase.org/11794
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 21, 2011
  2. unbreak automatic focusing of invalid field in phone home setup

    In addition to fixing it we're trying not to focus first invalid
    element if another invalid element is already focused. This sadly,
    doesn't work on IE8, seemingly because at form submit time focus is
    already on next button.
    
    Change-Id: I38d11934817a6a41fdc361e5822f03158959ed8e
    Reviewed-on: http://review.couchbase.org/11793
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 21, 2011
  3. fixed setup notifications errors display in IE8

    IE8 (quite correctly) doesn't like inserting block level elements into
    P elements. That's disallowed by html spec. In order to fix it we're
    changing element type to DIV and have to copy styles that were applied
    to P element in this place.
    
    Change-Id: I7a24a7dfff4adeac0e84a197d51bf6e83b97c153
    Reviewed-on: http://review.couchbase.org/11792
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 21, 2011
Commits on Dec 21, 2011
  1. wait for orchestrator presense for key operations. MB-4214 MB-4559

    Due to mb_master sometimes needing few seconds to detect mastership,
    we poll for readyness of orchestrator before doing key orchestrator
    calls.
    
    Change-Id: Icd267ba4551e833d8484dfdd71ec9c664660916a
    Reviewed-on: http://review.couchbase.org/11808
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk Nov 29, 2011
Commits on Dec 20, 2011
  1. initially hide (*) required field markers on registration form

    Change-Id: Iaf1eb7c4cd85b81f7e479bcc76a902b67e5473e4
    Reviewed-on: http://review.couchbase.org/11768
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 20, 2011
  2. implemented email validation for registration form

    Change-Id: I068f363b96f42515c1cfa792d49bf11a4de9e5e3
    Reviewed-on: http://review.couchbase.org/11762
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 20, 2011
  3. added HTML5 specified regexp for valid email addresses

    See
    http://www.w3.org/TR/html5/states-of-the-type-attribute.html#e-mail-state
    
    That's same regexp that html5 implementations are required to use for
    type=email inputs validation.
    
    Note: HTML5 explicitly and willfully violates some RFCs, but that
    still seems reasonable.
    
    Change-Id: I223ff61d007504ed5e7cb7f8f0c2f7153d8fcc51
    Reviewed-on: http://review.couchbase.org/11761
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 20, 2011
  4. expose observer callback

    So that we can force-run callback of any observer
    
    Change-Id: Ieef1dc8d467986d7b635ecdf446dc10d7e09681b
    Reviewed-on: http://review.couchbase.org/11760
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 20, 2011
  5. implemented 'plant my tree' 1.8 product registration changes

    This includes markup work from patch by Benjamin Young from here:
    http://review.couchbase.org/11696.
    
    And reverts not necessary change to set body class based on package
    edition. This: 9f78a8d. One of the
    issues was that it isn't compatible with forcing version through query
    parameter.
    
    Change-Id: I7e3f9ed478b1a1f113fa4e5fc629c3ca92620b25
    Reviewed-on: http://review.couchbase.org/11759
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 20, 2011
  6. don't preload images from no-preload directory

    Sometimes it's one-off images for setup wizard, for which there's no
    need to preload anything.
    
    Change-Id: I0a5d26d750e73d890cdbcb1499a65f5c4eefdfda
    Reviewed-on: http://review.couchbase.org/11758
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 19, 2011
  7. added mokugift tree image

    Change-Id: I3e110ee9fa67ff934d58923997dcfe75987239d4
    Reviewed-on: http://review.couchbase.org/11757
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 19, 2011
  8. initialize DAL.isEnterprise based on server's version

    Change-Id: I7a48dd976bfc284921f678f10154d48fe12f171e
    Reviewed-on: http://review.couchbase.org/11756
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 19, 2011
  9. amputate jquery's support for layer{X,Y} due to Chrome's warning

    Chrome developers decided to implemented extremely annoying warning
    for uses of layerX and layerY properties. We've never used them, but
    jquery's code to copy event properties from native event to jquery's
    emulation touch them. See
    http://code.google.com/p/chromium/issues/detail?id=101733
    
    Easy fix is to stop copying those.
    
    Change-Id: Iaad458c2ecbca44c1c9ac16232ed60f7e7e3df91
    Reviewed-on: http://review.couchbase.org/11755
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 19, 2011
  10. support forcing server version in UI

    By adding forceVersion query parameter to index.html
    (e.g. index.html?na&forceVersion=1.8.0-0-a-enterprise) it's now possible
    to override UI's idea of server version. So that testing
    community/enterprise-specific bits of UI is easy.
    
    Change-Id: Ie8035a15a66d0f5b924cd04d76c0d1dea507491d
    Reviewed-on: http://review.couchbase.org/11754
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 19, 2011
  11. avoid sending phone-home request on each entry into settings

    Recalculation of phEnabled cell caused phEnabled be undefined for a
    bit and then new values causes phEnabled observer to be always run
    regardless of actual changes to /settings/stats response.
    
    Change-Id: I67e71a1aeae3bdeb61a226d46e116f2ecf44037a
    Reviewed-on: http://review.couchbase.org/11771
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with alk Dec 19, 2011
  12. html-escape info field of phone home response

    Otherwise counterfeit DNS entry for phone-home domain may lead to
    XSS. This removes our ability to render HTML in info, but security
    risk of HTML in info is IMHO more important.
    
    Change-Id: I081b0ec563518a8c8605ec40e49a1fec934c97f5
    Reviewed-on: http://review.couchbase.org/11752
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Aliaksey Kandratsenka committed with steveyen Dec 19, 2011
  13. fixed highlighting of settings section

    Change-Id: I92b4275b7e87684b5f17cd9003fba7b55f35e212
    Reviewed-on: http://review.couchbase.org/11751
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Aliaksey Kandratsenka committed with steveyen Dec 19, 2011
  14. added enterprise|community classes based on version string

    Change-Id: I5fb3b9af915ff38cb8b7cb5a2443ecba573c6c89
    Reviewed-on: http://review.couchbase.org/11694
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    BigBlueHat committed with steveyen Dec 16, 2011
  15. generalized .invalid classes

    Change-Id: I441aa2996af34a50a0b77be7ddc69d9c5c0cdbb2
    Reviewed-on: http://review.couchbase.org/11695
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    BigBlueHat committed with alk Dec 16, 2011
Commits on Dec 17, 2011
  1. don't bother orchestrator when checking if rebalance is running

    Orchestrator call is potentially remote call. So doing it less often
    is good.
    
    Change-Id: I9617d292820ab9328b2e3eba62bdc9bbad0e5369
    Reviewed-on: http://review.couchbase.org/11680
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Aliaksey Kandratsenka committed with aartamonau Dec 15, 2011
Commits on Dec 15, 2011
  1. darken left nav headers. MB-4519

    Change-Id: I8ef1575a1366cc2269feab33d11daba35423e0c1
    Reviewed-on: http://review.couchbase.org/11678
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    BigBlueHat committed with alk Dec 15, 2011
Commits on Dec 13, 2011
  1. handle compatible mode heartbeats in normal mode

    In logs of MB-4507 we see mb_master crash when it receives compatible
    mode heartbeat with versioning mark from last node. It switches to
    normal mode, but node info is atom (not tuple) because node that sent
    heartbeat is still in compatible mode.
    
    We handle such hearbeat messages in same was as we do in compatible
    mode (comparing node names). Every other case is handled in same way
    as before.
    
    Change-Id: Ied4abf9ef4451f1b7d7c9c2da27698b118040d6e
    Reviewed-on: http://review.couchbase.org/11616
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Aliaksey Kandratsenka committed with aartamonau Dec 12, 2011
  2. handle undefined vbucket maps during failover. CBSE-40

    Particularly, it was possible to create bucket when one of nodes is
    down. That prevents janitor from assigning vbucket map. And this
    causes fail over code to see undefined vbucket map, which is handled
    now by doing nothing with such vbucket maps.
    
    Change-Id: Iee8c33abec6bd11513b4f9e4d0a48042f719ed84
    Reviewed-on: http://review.couchbase.org/11615
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Tested-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Reviewed-by: Aliaksey Artamonau <aliaksiej.artamonau@gmail.com>
    Aliaksey Kandratsenka committed with aartamonau Dec 12, 2011
Commits on Dec 12, 2011
  1. Rename MEMBASE_NUM_VBUCKETS to COUCHBASE_NUM_VBUCKETS.

    This is how this environment variable referred in README.markdown.
    
    Change-Id: I9bb3154d176c998b6281bcd67055e91f09dfe1ea
    Reviewed-on: http://review.couchbase.org/11561
    Tested-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    Reviewed-by: Aliaksey Kandratsenka <alkondratenko@gmail.com>
    aartamonau committed with alk Dec 8, 2011