lp:mysql-server/cluster-7.2

Created by Daniel Fischer on 2011-12-01 and last modified on 2015-04-17

NOTE: The MySQL Cluster repositories on Launchpad have been frozen. Future updates will be published on github instead.

Get this branch:
bzr branch lp:mysql-server/cluster-7.2
Members of Oracle/MySQL Engineering can upload to this branch. Log in for directions.

Branch merges

Related bugs

Related blueprints

Branch information

Owner:
Oracle/MySQL Engineering
Project:
MySQL Server
Status:
Mature

Recent revisions

4625. By Mauritz Sundell <email address hidden> on 2014-10-09

Cherrypicked

revision-id: <email address hidden>
parent: <email address hidden>
committer: Mauritz Sundell <email address hidden>
branch nick: mysql-7.1
timestamp: Thu 2014-10-09 14:46:36 +0200
message:
    Bug #19582807 MAKE SIGNAL DUMP IN TRACE FILES ALWAYS START FROM LATEST SIGNAL

    Removes a regression introduced with patch for the above bug.

    During crash dumps there could be a segmentation fault or most recent
    signals could be dump as old signals or not at all.

4624. By Mauritz Sundell <email address hidden> on 2014-10-06

Bug #17257842 EXCESSIVE RETRY FREQUENCY FOR API NODES RECONNECT CAUSES HIGH CPU ON MGM NODES.

Remove extra "ing" in new configuration parameters name to:

StartConnectBackoffMaxTime
ConnectBackoffMaxTime

4623. By Mauritz Sundell <email address hidden> on 2014-09-30

null merge 7.1 to 7.2

4622. By Balasubramanian Kandasamy on 2014-09-26

Added ndb_print_file man pages

4621. By Mauritz Sundell <email address hidden> on 2014-09-26

Bug #17257842 EXCESSIVE RETRY FREQUENCY FOR API NODES RECONNECT CAUSES HIGH CPU ON MGM NODES.

For system that have lot of API nodes and also have added data node
entries in configuration without starting any process for them the
API nodes will continously try to connect to the missing data nodes
several times per second. This put extra load on MGM nodes and
network in general.

This patch introduces logic to make an API node wait longer and
longer between attempts to connect to data nodes that fail to
respond.

The logic is controlled by two new parameters for api nodes:
StartingConnectingBackoffMaxTime and ConnectingBackoffMaxTime.

Both are given in milliseconds with approximately 100ms resolution.
Note that the time controls time *between* attempts, during the
time a connect attempt occur no time float with respect to the
backoff logic.

Between first and second attempt one waits 100ms (as before), but
for the following attempts the wait is doubled until the configures
maximal time are reached or ~100s whichever is smallest.

As long as the API node is not connected to any data node the
StartingConnectingBackoffMaxTime parameter is used, otherwise
ConnectingBackoffMaxTime.

Note, as soon as the API node is connected to a data node and that
node reports, in heartbeat message, that it has connected to some
other data nodes, connection attempts to those data nodes will not
be affected by the backoff logic but will be attempted every 100ms
until connected.

In worst case it can take HeartbeatIntervalDbApi (default 1,5s)
until the API node get notified that a restarting data node is up.

But since data nodes connects to each other during start phase 2,
and API node connections are allowed from start phase 8, that
should normally not be an actual delay since the time between start
phase 2 and 8 probably are at least several seconds.

For systems with many unstarted data nodes the new parameters
should be raised to avoid extreme traffic either to MGM nodes or
connection attempts to non existing datanodes.

4620. By Priyanka Sangam <email address hidden> on 2014-09-26

Bug#18487960: SIG 11 on delete with index merge

The mysqld cores while attempting read removal before a delete with
an index merge. The core occurs only if a quick select is generated
for the delete operation.

During read removal, the index from the quick select is used to index
into the table structure. But since the delete is an index merge, the
quick select index is set to MAX_KEY instead of a valid index value. So
the indexing operation generates a garbage pointer, which is then
dereferenced, resulting in a core.

Added a check so that read removal before delete is not attempted if
the quick select index has a value of MAX_KEY.

4619. By Mauritz Sundell <email address hidden> on 2014-09-24

Merge 5.5.40 into 7.2

4618. By saroj tripathy <email address hidden> on 2014-09-24

merge 7.1 -> 7.2

4617. By Pratik Patodi<email address hidden> on 2014-09-23

merge 7.1->7.2

4616. By Magnus BlÄudd on 2014-09-18

Bug #19365180 PORT FIX FOR BUG 17283409 TO MYSQL CLUSTER 7.2

 - add temp MCP hack to avoid DEBUG_SYNC() when current_thd is NULL
 - current_thd is NULL when function is called for example during startup

Branch metadata

Branch format:
Branch format 7
Repository format:
Bazaar repository format 2a (needs bzr 1.16 or later)
Stacked on:
lp:~dannythefool/+junk/trunk-outdated
This branch contains Public information 
Everyone can see this information.