Revision tags: v4.6.5, v4.6.3, v4.6.4, v4.6.2, v4.6.1, v4.6.0, v4.6.0-DP, v4.5.1, v4.1.2, v3.1.6, v4.5.0, v4.1.1, v3.1.5, v3.1.4, v3.1.3, v4.1.0, v3.1.2, v3.1.1, v4.0.0, v3.1.0, v3.0.2 |
|
#
1e3ce0f6 |
| 11-Sep-2014 |
Mike Wiederhold <mike@couchbase.com> |
Fix deadlock in checkpoint persistence command We need to release the hpChkMutex before notifying memcached in order to prevent a deadlock. The deadlock occurs when the flusher trys
Fix deadlock in checkpoint persistence command We need to release the hpChkMutex before notifying memcached in order to prevent a deadlock. The deadlock occurs when the flusher trys to notify memcached of checkpoint persistence at the same time a memcached worker thread is trying to add a new checkpoint persistence request. Change-Id: Ida313f5b39ef0e063dee9882410cd0a19ce55292 Reviewed-on: http://review.couchbase.org/41363 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Michael Wiederhold <mike@couchbase.com>
show more ...
|
Revision tags: v3.0.0 |
|
#
91d3524d |
| 12-Aug-2014 |
Sriram Ganesan <sriram@couchbase.com> |
Add id to the logs of checkpoint persistence Add the id to the logs of checkpoint persistence so that we will have the sequence number/checkpoint id in case a timeout occurs on checkpoin
Add id to the logs of checkpoint persistence Add the id to the logs of checkpoint persistence so that we will have the sequence number/checkpoint id in case a timeout occurs on checkpoint persistence Change-Id: Ifd3c70c9d40de43f0b3f1fdc1cf495ee0dc95d60 Reviewed-on: http://review.couchbase.org/40513 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Sriram Ganesan <sriram@couchbase.com>
show more ...
|
#
007528bb |
| 07-Aug-2014 |
Mike Wiederhold <mike@couchbase.com> |
MB-11887: Properly send the snapshot start and end seqnos Currently we always set the snapshot start and end seqno to be the same as the start seqno. This is incorrect if we recently wer
MB-11887: Properly send the snapshot start and end seqnos Currently we always set the snapshot start and end seqno to be the same as the start seqno. This is incorrect if we recently were disconnected and need to resume from a partial snapshot. Change-Id: I7c9430e0b24ef01cc427cb03b518f4c41a55d875 Reviewed-on: http://review.couchbase.org/40276 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Michael Wiederhold <mike@couchbase.com>
show more ...
|
#
b931e237 |
| 31-Jul-2014 |
Sundar Sridharan <sundar.sridharan@gmail.com> |
fix compiler error in vbucket.cc metaDataDisk += Change-Id: I5b2992d3298d5b65f959f48cb078ed1c1f7a01b1 Reviewed-on: http://review.couchbase.org/40150 Reviewed-by: abhinav dangeti <abh
fix compiler error in vbucket.cc metaDataDisk += Change-Id: I5b2992d3298d5b65f959f48cb078ed1c1f7a01b1 Reviewed-on: http://review.couchbase.org/40150 Reviewed-by: abhinav dangeti <abhinav@couchbase.com> Tested-by: Sundararaman Sridharan <sundar@couchbase.com>
show more ...
|
#
b49c193e |
| 30-Jul-2014 |
D Liao <liaodw@gmail.com> |
MB-11689 Added a stat ep_meta_data_disk This gives approx disk space used to store item meta data by couchstore. Change-Id: I7f1642279c9aca84b02c885fd621cfdf50c20e98 Reviewed-on
MB-11689 Added a stat ep_meta_data_disk This gives approx disk space used to store item meta data by couchstore. Change-Id: I7f1642279c9aca84b02c885fd621cfdf50c20e98 Reviewed-on: http://review.couchbase.org/40038 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
#
209356e3 |
| 29-Jul-2014 |
abhinavdangeti <abhinav@couchbase.com> |
MB-11629: Set engineSpecific to NULL when notify_failed_conns Context: While notifying pending connections failed Change-Id: I5e4085794a70a068f7a95735242b1e841cc6cf8f Reviewed-o
MB-11629: Set engineSpecific to NULL when notify_failed_conns Context: While notifying pending connections failed Change-Id: I5e4085794a70a068f7a95735242b1e841cc6cf8f Reviewed-on: http://review.couchbase.org/39982 Reviewed-by: abhinav dangeti <abhinav@couchbase.com> Tested-by: abhinav dangeti <abhinav@couchbase.com>
show more ...
|
#
88f0d4d5 |
| 28-Jul-2014 |
Mike Wiederhold <mike@couchbase.com> |
Add vbucket-seqno stats to vbucket-details We should include these stats in the vbucket details stats as well since vbucket-seqno is really only mean to be used internally for client
Add vbucket-seqno stats to vbucket-details We should include these stats in the vbucket details stats as well since vbucket-seqno is really only mean to be used internally for clients that only want to get the bare minimum needed to make dcp connections. Change-Id: Ib95b80f8ec7afb37b146d6932e7c722da9d917c4 Reviewed-on: http://review.couchbase.org/39968 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Reviewed-by: Sundararaman Sridharan <sundar@couchbase.com> Tested-by: Michael Wiederhold <mike@couchbase.com>
show more ...
|
#
134a9baf |
| 15-May-2014 |
Chiyoung Seo <chiyoung.seo@gmail.com> |
MB-11026 Don't allow the underflow on the dirty queue size stat. We will find the root cause of the underflow issue separately soon. Change-Id: Id5459fe3ac14f850159e932179113786
MB-11026 Don't allow the underflow on the dirty queue size stat. We will find the root cause of the underflow issue separately soon. Change-Id: Id5459fe3ac14f850159e9321791137860f265f62 Reviewed-on: http://review.couchbase.org/37153 Reviewed-by: abhinav dangeti <abhinav@couchbase.com> Tested-by: abhinav dangeti <abhinav@couchbase.com>
show more ...
|
#
c649b2d9 |
| 17-Mar-2014 |
Trond Norbye <trond.norbye@gmail.com> |
MB-10476: Use cb_assert instead of assert If compiled with -DNDEBUG assert is an empty macro Change-Id: I266ddf8a8fe3f1a388a9d8ce3d21f28614c35315 Reviewed-on: http://review.couc
MB-10476: Use cb_assert instead of assert If compiled with -DNDEBUG assert is an empty macro Change-Id: I266ddf8a8fe3f1a388a9d8ce3d21f28614c35315 Reviewed-on: http://review.couchbase.org/34579 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
#
15aed944 |
| 11-Mar-2014 |
Mike Wiederhold <mike@couchbase.com> |
Reduce the amount of headers that are imported in .h files This change removes imports of certain header files from all other header files and into .cc files. The problem with import
Reduce the amount of headers that are imported in .h files This change removes imports of certain header files from all other header files and into .cc files. The problem with importing a header in another header file is that any change causes all headers used in that header file to be recompiled. In some cases changing one header file causes us to have to recompile everyhting. This change moves some of these headers to only .cc files and should reduce compile time. File no longer imported from other headers: conflict_resolution.h failover-table.h flusher.h mutation_log.h tap-throttle.h upr-consumer.h upr-consumer.h Change-Id: Idd44bcaac874e294c1085da46760badf37a1ea05 Reviewed-on: http://review.couchbase.org/34395 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
Revision tags: v2.5.1 |
|
#
53e8562f |
| 29-Jan-2014 |
Chiyoung Seo <chiyoung.seo@gmail.com> |
Track the number of items per vbucket with the full eviction. As the full eviction ejects an item from cache completely, we need a new way of tracking the number of item per vbucket when
Track the number of items per vbucket with the full eviction. As the full eviction ejects an item from cache completely, we need a new way of tracking the number of item per vbucket when the full eviction is chosen. This change basically adds a new flag to the hash table item (i.e., StoredValue) to indicate whether an item is newly added to the cache or not, and then increments the item counter. Then, the flusher figures out whether the counter should be decremented because the item already exists on the DB file. Change-Id: I492c03053238a4baa155eb76575893da323d26c4 Reviewed-on: http://review.couchbase.org/32985 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
#
1ba4011d |
| 18-Jan-2014 |
Sundar Sridharan <sundar.sridharan@gmail.com> |
MB-9919: Provide DB data size and file size stats This change provides DB data size and file size per vbucket and their aggregated values through stats call. stats "vbucket-details v
MB-9919: Provide DB data size and file size stats This change provides DB data size and file size per vbucket and their aggregated values through stats call. stats "vbucket-details vbucket_id" returns the list of a given vbucket's stats including its DB data size and file size. For example, "vbucket-details 0" includes the following stats for vbucket 0: 1) vb_0:db_data_size --> DB data size for vbucket 0 2) vb_0:db_file_size --> DB file size for vbucket 0 On the other hand, stats "all" includes the aggregated DB data size and file size: 1) ep_db_data_size --> DB data size aggreaged across all vbuckets. 2) ep_db_file_size --> DB file size aggregated across all vbuckets. A test case is added to ensure the value ticks up correctly. Change-Id: Id8ae17ae7726f9a003a3245defc2ccf4ee89b77c Reviewed-on: http://review.couchbase.org/32533 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
#
18770839 |
| 07-Jan-2014 |
Sundar Sridharan <sundar.sridharan@gmail.com> |
coding style fixes for gerrit compatibility gerrit does not handle lines more than 80 characters well Also coding style dictates fixed line sizes which are mostly followed but left o
coding style fixes for gerrit compatibility gerrit does not handle lines more than 80 characters well Also coding style dictates fixed line sizes which are mostly followed but left out in a few places.. - remove trailing whitespaces from all files - limit lines to 80 chars - try to ensure comments fit into one line so that they can be traced back from an error log Change-Id: Id8f1c34c3c7cb698a1b50f5f225b4516667f5e42 Reviewed-on: http://review.couchbase.org/31639 Tested-by: Sundararaman Sridharan <sundar.sridharan@gmail.com> Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
#
9d8a9725 |
| 23-Dec-2013 |
Trond Norbye <trond.norbye@gmail.com> |
Move towards using C++11 atomics Start by renamimg all use of Atomics that doesn't "break" by switching to std::atomic to AtomicValue (when the transition is done we'll rename everyt
Move towards using C++11 atomics Start by renamimg all use of Atomics that doesn't "break" by switching to std::atomic to AtomicValue (when the transition is done we'll rename everything back). Change-Id: I3f694856dab498c565806c9ac1ee5baf3bc07b17 Reviewed-on: http://review.couchbase.org/31333 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
#
a00c1a75 |
| 23-Dec-2013 |
David Liao <liaodw@gmail.com> |
MB-9741 Replace usage of QueuedItem with Item class. This is the first phase of reducing memory used by keys/metadata of items. This checkin now replace QueuedItem with Item used by Chec
MB-9741 Replace usage of QueuedItem with Item class. This is the first phase of reducing memory used by keys/metadata of items. This checkin now replace QueuedItem with Item used by Checkpoing Manager and prepares for using Item class in StoredValue. QueuedItem is removed from the code base. Change-Id: I6afecac51410f209155cf357f6a276c20670b026 Reviewed-on: http://review.couchbase.org/31050 Reviewed-by: Michael Wiederhold <mike@couchbase.com> Tested-by: Michael Wiederhold <mike@couchbase.com>
show more ...
|
#
79a9d87d |
| 19-Dec-2013 |
David Liao <liaodw@gmail.com> |
MB-9638 UPR persistence sequence number command This command is similar to the Tap checkpoint persistence command. But we use sequence number instead of checkpoint id for UPR. Refact
MB-9638 UPR persistence sequence number command This command is similar to the Tap checkpoint persistence command. But we use sequence number instead of checkpoint id for UPR. Refactored code to support both checkpoint id and seqno persistence. The new memcached command opcode is CMD_SEQNO_PERSISTENCE 0xb7. Added a stat "persistence_seqno". Added a test case "upr_persistence_seqno". Change-Id: Iec5db260e05cfa8889f253c33bb08209eb87ca2c Reviewed-on: http://review.couchbase.org/30899 Reviewed-by: Michael Wiederhold <mike@couchbase.com> Tested-by: Michael Wiederhold <mike@couchbase.com>
show more ...
|
#
7c2254cd |
| 27-Nov-2013 |
Chiyoung Seo <chiyoung.seo@gmail.com> |
MB-9161 Reset engine_specific to NULL when chk_persistence fails If a checkpoint_persistence request fails due to the timeout, we should then reset the engine_specific value to NULL, so
MB-9161 Reset engine_specific to NULL when chk_persistence fails If a checkpoint_persistence request fails due to the timeout, we should then reset the engine_specific value to NULL, so that the next checkpoint persistence request should work correctly. Change-Id: Iab0ab3c41fe39dc7060c3d094ec04cb9cfadcb73 Reviewed-on: http://review.couchbase.org/30677 Reviewed-by: Sundararaman Sridharan <sundar.sridharan@gmail.com> Tested-by: Sundararaman Sridharan <sundar.sridharan@gmail.com>
show more ...
|
#
7c6809d2 |
| 07-Nov-2013 |
Trond Norbye <trond.norbye@gmail.com> |
Start migrate the API for our Atomics to look like std::atomic This patch change the Atomic class CouchbaseAtomic, and change the API so that: get() => load() set() => store(
Start migrate the API for our Atomics to look like std::atomic This patch change the Atomic class CouchbaseAtomic, and change the API so that: get() => load() set() => store() cas() => compare_exchange_strong() and move some of methods that used to be members as template functions. Change-Id: I48d873fbf04f40d446444232db44f565ee4fed54 Reviewed-on: http://review.couchbase.org/30135 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
#
92cd2ded |
| 16-Nov-2013 |
Mike Wiederhold <mike@couchbase.com> |
MB-9311: Notify all the pending connections on vbucket delete If a given vbucket is deleted, we should notify it to all of its pending connections, so that the connections are released
MB-9311: Notify all the pending connections on vbucket delete If a given vbucket is deleted, we should notify it to all of its pending connections, so that the connections are released immediately and the error messages should be sent to the clients. Change-Id: Ic230a803f0b26944f98e84ed4d21f8d834e3a561 Reviewed-on: http://review.couchbase.org/30353 Reviewed-by: Michael Wiederhold <mike@couchbase.com> Tested-by: Michael Wiederhold <mike@couchbase.com>
show more ...
|
#
a4d46f6b |
| 01-Jul-2013 |
Mike Wiederhold <mike@couchbase.com> |
MB-8988: Sequence numbers are assigned by the CheckpointManager Change-Id: If3fdf08e1091ad36ea8bcd30b10cd3803db009ae Reviewed-on: http://review.couchbase.org/29226 Reviewed-by: Chiyo
MB-8988: Sequence numbers are assigned by the CheckpointManager Change-Id: If3fdf08e1091ad36ea8bcd30b10cd3803db009ae Reviewed-on: http://review.couchbase.org/29226 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
#
9a8679e2 |
| 17-Oct-2013 |
Chiyoung Seo <chiyoung.seo@gmail.com> |
CBD-1034 Support the full eviction as an optional eviction policy When an item is evicted from the cache, its value is only evicted from memory by default. To further reduce the memory
CBD-1034 Support the full eviction as an optional eviction policy When an item is evicted from the cache, its value is only evicted from memory by default. To further reduce the memory overhead, this change supports the full eviction as an additional option, which removes an item's key, metadata, and value from the cache. The eviction policy can be chosen when a bucket is instantiated. If the full eviction option is chosen, this results in affecting lots of engine APIs and also causes more disk IO overhead. For example, CAS for a non-resident item will require a backgroud fetch because it should know the item's CAS value. Change-Id: I2edbe2b12527ef1c1fb31992cb04afbb7c078dac Reviewed-on: http://review.couchbase.org/29155 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
#
eedbb129 |
| 25-Sep-2013 |
Chiyoung Seo <chiyoung.seo@gmail.com> |
MB-9159 Batch read uses a key instead of a sequence number. When an item is evicted from memory, we plan to evict all the metadata fields including the item's sequence number in addiiton
MB-9159 Batch read uses a key instead of a sequence number. When an item is evicted from memory, we plan to evict all the metadata fields including the item's sequence number in addiiton to the item's value. This new eviction approach requires a batch read to use a key instead of a sequence number when it reads items from disk. This change adapts the batch read, so that it can use a key to fetch an item's metadata or value from disk. Change-Id: Ief14fd0ef4f42b316e75c6719349cdb1c63fac80 Reviewed-on: http://review.couchbase.org/29181 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
#
f7806180 |
| 30-Oct-2013 |
Sundar Sridharan <sundar.sridharan@gmail.com> |
MB-9434: prevent deadlock between flusher & stats If the memcached stats thread which is holding the thread lock tries to acquire the high priority checkpoint list lock just to read
MB-9434: prevent deadlock between flusher & stats If the memcached stats thread which is holding the thread lock tries to acquire the high priority checkpoint list lock just to read its size, it can deadlock on the flusher thread which may be holding the list lock & trying to notify the same memcached thread doing the stats To avoid this deadlock maintain the size of the high priority checkpoints in a separate variable and allow for dirty read on the size Change-Id: I6f0991c22deb2cc7762e5967496b99c3baed2c12 Reviewed-on: http://review.couchbase.org/29900 Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com> Tested-by: Chiyoung Seo <chiyoung@couchbase.com>
show more ...
|
#
51a0f7d9 |
| 22-Oct-2013 |
abhinavdangeti <abhinav@couchbase.com> |
MB-9098: Acquire lock to retrieve list size - Need for synchronizing access to shared data to avoid segmentation faults and possible memory corruption. - Backporting from ma
MB-9098: Acquire lock to retrieve list size - Need for synchronizing access to shared data to avoid segmentation faults and possible memory corruption. - Backporting from master: http://review.couchbase.org/#/c/29477/ Change-Id: Id24770dbc72c082967959aee30c8e21b9556989e Reviewed-on: http://review.couchbase.org/29687 Reviewed-by: Sundararaman Sridharan <sundar.sridharan@gmail.com> Tested-by: Sundararaman Sridharan <sundar.sridharan@gmail.com>
show more ...
|
#
acb436bb |
| 14-Oct-2013 |
abhinavdangeti <abhinav@couchbase.com> |
MB-9098: Acquire lock to retrieve list size - Need for synchronizing access to shared data to avoid segmentation faults and possible memory corruption. Change-Id: If02dcf0d6
MB-9098: Acquire lock to retrieve list size - Need for synchronizing access to shared data to avoid segmentation faults and possible memory corruption. Change-Id: If02dcf0d61a397960f0e7bd0b738fe1bb5f93cbb Reviewed-on: http://review.couchbase.org/29477 Reviewed-by: abhinav dangeti <abhinav.dangeti@gmail.com> Tested-by: abhinav dangeti <abhinav.dangeti@gmail.com>
show more ...
|