Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Commits on Mar 4, 2015
  1. abhinav dangeti

    Merge remote-tracking branch 'gerrit/sherlock'

    abhinavdangeti authored
    Change-Id: Ibdbbd9f4dddcace840a613ffb3f087282c058f56
  2. abhinav dangeti

    MB-13155: Decrement dirtyQueue size after rollback

    abhinavdangeti authored
    We will need to decrement the dirty queue size stat
    of the vbucket whose checkpoints have been cleared
    during rollback.
    
    Change-Id: I06fc32ac8d7ab77b1bc47df51b3809aab728dce8
    Reviewed-on: http://review.couchbase.org/47580
    Reviewed-by: Manu Dhundi <manu@couchbase.com>
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Tested-by: abhinav dangeti <abhinav@couchbase.com>
Commits on Mar 3, 2015
  1. MB-13568: notify the flusher if items are present in the reject queue

    Sriram Ganesan authored
    If items are present in the reject queue and if the flusher is not
    running, then there is a possibility that those items will be
    stuck in the diskwrite queue until the flusher gets notified again.
    
    Change-Id: Iaeff78b5700ea5e7fe27cf186173ec82e7a00c82
    Reviewed-on: http://review.couchbase.org/47592
    Reviewed-by: abhinav dangeti <abhinav@couchbase.com>
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Tested-by: Sriram Ganesan <sriram@couchbase.com>
Commits on Feb 27, 2015
  1. Mike Wiederhold

    [BP] MB-13356: If the buffer is full check to see if we have dead bac…

    mikewied authored mikewied committed
    …kfills
    
    It's possible that the buffer becomes full because a backfill is still
    running for a dead vbucket. If this is the case remove the backfill
    from the backfill manager.
    
    Change-Id: Id681fb0e1a7cf93aa07325664e9dc8709f3d7650
    Reviewed-on: http://review.couchbase.org/47463
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
  2. Mike Wiederhold

    [BP] Add a move to active queue function to simplify the backfill fun…

    mikewied authored mikewied committed
    …ction
    
    Change-Id: Iad559ead8bb2d7d9f5d2afdc6b38d334fa2f277b
    Reviewed-on: http://review.couchbase.org/47462
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
  3. Mike Wiederhold

    [BP] Remove redundent check for buffer being full after a backfill

    mikewied authored mikewied committed
    This check is redundent because we can just run the backfill
    manager task one more time and it will pause the backfill manager
    then.
    
    Change-Id: I069e5e273167c1d3ba85f1e6bb0dfc876fa878e3
    Reviewed-on: http://review.couchbase.org/47461
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
  4. Mike Wiederhold

    [BP] If a backfill task is finished cancel should return immediately

    mikewied authored mikewied committed
    This will prevent a crash from happening when trying to free up
    resources for the backfill a second time.
    
    Change-Id: I2625d3009291de01fe01cdf3ff7d08c10b800e60
    Reviewed-on: http://review.couchbase.org/47460
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
Commits on Feb 26, 2015
  1. Mike Wiederhold

    MB-13356: If the buffer is full check to see if we have dead backfills

    mikewied authored mikewied committed
    It's possible that the buffer becomes full because a backfill is still
    running for a dead vbucket. If this is the case remove the backfill
    from the backfill manager.
    
    Change-Id: I5c248d71d79dd9fd1796d5b39ed714d8253dc410
    Reviewed-on: http://review.couchbase.org/47227
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Reviewed-by: Manu Dhundi <manu@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
  2. Mike Wiederhold

    Add a move to active queue function to simplify the backfill function

    mikewied authored mikewied committed
    Change-Id: Ic709e0e23604041cf1f7ec90ae0ad065bcacd30a
    Reviewed-on: http://review.couchbase.org/47297
    Reviewed-by: Manu Dhundi <manu@couchbase.com>
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
  3. Mike Wiederhold

    Remove redundent check for buffer being full after a backfill

    mikewied authored mikewied committed
    This check is redundent because we can just run the backfill
    manager task one more time and it will pause the backfill manager
    then.
    
    Change-Id: Ic627be7850ce6556f31b097e918aa121f43b40e4
    Reviewed-on: http://review.couchbase.org/47238
    Reviewed-by: Manu Dhundi <manu@couchbase.com>
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
  4. Mike Wiederhold

    If a backfill task is finished cancel should return immediately

    mikewied authored mikewied committed
    This will prevent a crash from happening when trying to free up
    resources for the backfill a second time.
    
    Change-Id: I2a260942161d1e2445487e2afc96d82a4398982d
    Reviewed-on: http://review.couchbase.org/47419
    Reviewed-by: Manu Dhundi <manu@couchbase.com>
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
Commits on Feb 25, 2015
  1. Dave Rigby Chiyoung Seo

    MB-13591: Defragmenter: ensure time_per_item != 0

    daverigby authored chiyoung committed
    During defragmentation we calculate the rate at which items have been
    visited, to be able to determine when to pause the defragmenter
    task.
    
    However there was a potential divide-by-zero error when calcuating the
    time_per_item; if we calculate this as less than one nanosecond it
    results in a value of zero, which is used later as divisor.  While one
    would assume it would be impossible to process an item in less than
    one nanosecond, due to timer precision we may get such a result.
    
    Solution is to add on the timer period to time_per_item; which will
    ensure that it is always non-zero.
    
    Change-Id: I6ae26d858c4b4ef4ea4bcc5bd3a108ce18b73268
    Reviewed-on: http://review.couchbase.org/47343
    Reviewed-by: Trond Norbye <trond.norbye@gmail.com>
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
Commits on Feb 24, 2015
  1. abhinav dangeti

    Merge remote-tracking branch 'gerrit/3.0.3' into master

    abhinavdangeti authored
    + MB-13386: Ensure that purging the highSeqno doesn't happen
    
    Change-Id: I7c9e33f94fe889e666268e38bd5432aa978a557b
  2. abhinav dangeti

    Merge remote-tracking branch 'gerrit/3.0.1' into 3.0.3

    abhinavdangeti authored
    + MB-13386: Ensure that purging the highSeqno doesn't happen
    
    Change-Id: Ieacb1d86d747a275181bb238145fd206a14be140
  3. abhinav dangeti

    MB-13386: Ensure that purging the highSeqno doesn't happen

    abhinavdangeti authored
    + The highest seqno is to not be purged for DCP.
    + This change ensures that items are still queued
    for deletion if found to be expired.
    
    Change-Id: I8102d5f61989523efc4f39b70b225c05cdd1b128
    Reviewed-on: http://review.couchbase.org/47241
    Reviewed-by: Manu Dhundi <manu@couchbase.com>
    Reviewed-by: Michael Wiederhold <mike@couchbase.com>
    Tested-by: abhinav dangeti <abhinav@couchbase.com>
  4. MB-7965: Skip scanning for existing files during flush all

    Sriram Ganesan authored
    During flush all, all the vbucket files are being deleted
    and recreated. Given that the files are going to be
    recreated anyways, there is no need to scan the directory
    for any existing files.
    
    Change-Id: I9b6e0c39f193e282441aba6e481cc1ebbff1a64c
    Reviewed-on: http://review.couchbase.org/47134
    Reviewed-by: abhinav dangeti <abhinav@couchbase.com>
    Tested-by: Sriram Ganesan <sriram@couchbase.com>
Commits on Feb 23, 2015
  1. Mike Wiederhold

    Use backfill buffered items for the backfill buffered items stat

    mikewied authored mikewied committed
    Change-Id: Ia9401b869fa314210575cbe61e0da44889416824
    Reviewed-on: http://review.couchbase.org/47225
    Reviewed-by: abhinav dangeti <abhinav@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
  2. Mike Wiederhold

    MB-13359: Get DCP next snapshot if there are meta items, but no mutat…

    mikewied authored mikewied committed
    …ions
    
    If we only get a snapshot end message when we call getItemsForCursor
    then we will consider the snapshot to be empty and pause the stream.
    Since a snapshot end message can only be in a closed checkpoint this
    means that it's possible that we are pausing the stream for no
    reason.
    
    Change-Id: I29b8603287b41401fd6f5c1e4d4f185611d5b583
    Reviewed-on: http://review.couchbase.org/47087
    Reviewed-by: abhinav dangeti <abhinav@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
  3. Mike Wiederhold

    MB-13356: Use a list for all backfill queues

    mikewied authored mikewied committed
    We want to be able to iterate these so we should use a list instead
    of a queue.
    
    Change-Id: Ie8fa1ece137586a47de82ee233acfda5393d48df
    Reviewed-on: http://review.couchbase.org/47226
    Reviewed-by: Manu Dhundi <manu@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
Commits on Feb 20, 2015
  1. abhinav dangeti

    Set nvalue to 1 before calling get_item_info

    abhinavdangeti authored
    Change-Id: I3e48586ca43b7bace879618446bba59eb5ab81af
    Reviewed-on: http://review.couchbase.org/47133
    Reviewed-by: Sriram Ganesan <sriram@couchbase.com>
    Reviewed-by: Manu Dhundi <manu@couchbase.com>
    Tested-by: abhinav dangeti <abhinav@couchbase.com>
  2. abhinav dangeti

    Merge remote-tracking branch 'gerrit/3.0.3'

    abhinavdangeti authored
    Change-Id: Id9b426d6e209ce40ae72c34d2b55f234b9b0bbbf
  3. abhinav dangeti

    Fix build breakage on ubuntu

    abhinavdangeti authored
    Change-Id: Ic8d7b9c17086a612b55e2c84a26259081d29c8d4
    Reviewed-on: http://review.couchbase.org/47107
    Reviewed-by: Manu Dhundi <manu@couchbase.com>
    Tested-by: abhinav dangeti <abhinav@couchbase.com>
  4. manudhundi

    Merge remote-tracking branch 'couchbase/3.0.3'

    manudhundi authored
    Change-Id: I4f2ab37bd02943a3ad0972f0a49fcebe80fbbfaa
  5. manudhundi

    MB-13479: Indicate rollback to DCP clients if purge_seqno > snap_star…

    manudhundi authored
    …t_seqno
    
    The replica may not get all the items if there are purged items in the active
    node. Hence this solution proposes to indicate a rollback to seqno 0 in case
    the purge_seqno > snap_start_seqno.
    Note: Since the occurrence of this scenario is pretty rare, rolling back to 0
    may not be a bad thing.
    
    Change-Id: I5c8403115110be136df5d4cb4e2704edc2a4c9e4
    Reviewed-on: http://review.couchbase.org/47094
    Tested-by: Manu Dhundi <manu@couchbase.com>
    Reviewed-by: Michael Wiederhold <mike@couchbase.com>
Commits on Feb 19, 2015
  1. Mike Wiederhold

    MB-13356: If backfill remaining is 0 we still need to decr the buffer…

    mikewied authored mikewied committed
    … size
    
    Change-Id: I4729ee46f316cb0c97d7e347e1797e1dc839bd72
    Reviewed-on: http://review.couchbase.org/47086
    Reviewed-by: abhinav dangeti <abhinav@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
  2. Mike Wiederhold

    MB-13356: Change the name of the stream buffer stats for consistency

    mikewied authored mikewied committed
    Changing the name from buf_backfill to backfill_buffer.
    
    Change-Id: Ib42f5d43f51c3f6f70bfd4a642d4c56d0babb947
    Reviewed-on: http://review.couchbase.org/47085
    Reviewed-by: abhinav dangeti <abhinav@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
  3. Mike Wiederhold

    MB-13356: Add stats for the backfill manager buffer

    mikewied authored mikewied committed
    These stats can be used to debug issues caused by incorrect buffer
    accounting.
    
    Change-Id: Ic6a8c4c01d9b19dfe935e165aefeec9aede621cc
    Reviewed-on: http://review.couchbase.org/47084
    Reviewed-by: abhinav dangeti <abhinav@couchbase.com>
    Tested-by: Michael Wiederhold <mike@couchbase.com>
  4. manudhundi

    Merge remote-tracking branch 'couchbase/3.0.3' to 'couchbase/master'

    manudhundi authored
    Change-Id: Ibc0ad49d81167ce4a187f6cff12aa90c506f5fcc
  5. abhinav dangeti

    MB-13517: Initialize numActiveSnoozingBackfills to zero

    abhinavdangeti authored
    The failure to initialize this parameter would sometimes
    give it a garbage value, which would cause the check of
    the current active and snoozing backfill count against
    the threshold value to fail, thereby backfills will never
    be added into the active queue.
    
    Change-Id: Ic8d6f7c06e9067dff35a678ba51d35336d29e082
    Reviewed-on: http://review.couchbase.org/47043
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Reviewed-by: Michael Wiederhold <mike@couchbase.com>
    Tested-by: abhinav dangeti <abhinav@couchbase.com>
  6. manudhundi

    Merge remote-tracking branch 'couchbase/3.0.1' into couchbase/3.0.3

    manudhundi authored
    Change-Id: I3d214e5b9b7a60831dcb685a6ef85a4ee0996f3e
  7. manudhundi

    MB-13386: Do not purge the item with highest sequence number from db.

    manudhundi authored
    When the highest seq number that DCP is supposed to read from
    the db is the last seq number in the db and is purged,
    DCP backfill does not know about it and hence waits for it.
    This results in DCP connection hang.
    To solve this problem, during compaction we do not purge the
    item with last(highest) sequence number in the db.
    
    Change-Id: Ib83c335da3f7c0a952e4b760309276f73bff4ccf
    Reviewed-on: http://review.couchbase.org/47035
    Reviewed-by: Michael Wiederhold <mike@couchbase.com>
    Tested-by: Manu Dhundi <manu@couchbase.com>
Commits on Feb 18, 2015
  1. abhinav dangeti

    Merge remote-tracking branch 'gerrit/3.0.3' into MASTER

    abhinavdangeti authored
    MB-10370: Replications' set/delWithMetas to use replication threshold
    
    Change-Id: Ifa1c004bcc32ac0603ce2ef537c5e6226acec827
  2. abhinav dangeti

    Fix compilation warning in checkpoint.cc

    abhinavdangeti authored
    Change-Id: I320f81465abb1c5e6c0017987987bd6d43b325ba
    Reviewed-on: http://review.couchbase.org/46991
    Reviewed-by: abhinav dangeti <abhinav@couchbase.com>
    Tested-by: abhinav dangeti <abhinav@couchbase.com>
  3. Update the flexible metadata length correctly during retrieval

    Sriram Ganesan authored
    While retrieving the item from couchstore, ensure that the flexible
    metadata length is updated correctly. Without this update, the
    flexible metadata length had garbage values resulting in data
    corruption.
    
    Change-Id: I4fba3047d755c3000fc9b3aa2d46dc5271119d91
    Reviewed-on: http://review.couchbase.org/46928
    Reviewed-by: abhinav dangeti <abhinav@couchbase.com>
    Tested-by: Sriram Ganesan <sriram@couchbase.com>
Commits on Feb 17, 2015
  1. abhinav dangeti Chiyoung Seo

    MB-10370: Replications' set/delWithMetas to use replication threshold

    abhinavdangeti authored chiyoung committed
    The setWithMetas and deleteWithMetas issue by consumers for
    intra-cluster replication will need to work on tapThrottleThreshold
    as opposed to mutation_memory_threshold.
    
    Change-Id: I576c9f9961e03e70430d58a192854c6faa14156d
    Reviewed-on: http://review.couchbase.org/46907
    Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
    Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
Something went wrong with that request. Please try again.