Timestamps are different
I am in the very early stages of testing out our Galara cluster with innodb databases.
I have a tool that is comparing data against each node on the cluster and I have found that timestamps seem to be the only thing that is inconstant and by exactly 5 hours.
I have a record in a Galera cluster that has a field with date time.
I took one database and started comparing each table (one by one). I had several thousand records with different timestamps. An example is 2013-07-03 11:13:59, but another server, the date time is 2013-07-03 16:13:59. I would suspect that the times would be synced everywhere exactly the same.
If I run a "SELECT NOW();" on all four servers at the same time, I see: +---------------------+
| NOW() |
+---------------------+
| 2013-08-13 14:59:51 |
+---------------------+ +---------------------+
| NOW() |
+---------------------+
| 2013-08-13 14:59:51 |
+---------------------+ +---------------------+
| NOW() |
+---------------------+
| 2013-08-13 14:59:51 |
+---------------------+ +---------------------+
| NOW() |
+---------------------+
| 2013-08-13 14:59:51 |
+---------------------+
If I run a select @@global.time_zone, @@session.time_zone on all four nodes in the cluster, I also see +--------------------+---------------------+
| @@global.time_zone | @@session.time_zone |
+--------------------+---------------------+
| SYSTEM | SYSTEM |
+--------------------+---------------------+
And if I enter date on all 4 machines, the time matches just fine.
I have checked other non erip databases (some very large databases) and things are correct within those systems.
[root@a2 ]# mysql -u root -p** -h a2 -e "select * from safeplan_new1.plan_approval where plan_approval_id=3247" +------------------+----------------+------------------+------------------------+------------------+---------------------+----------------------+
| plan_approval_id | plan_master_id | plan_location_id | plan_approval_location | plan_approval_by | plan_approved_on | plan_approval_status |
+------------------+----------------+------------------+------------------------+------------------+---------------------+----------------------+
| 3247 | 292 | 0 | 155 | 133 | 2013-07-03 11:13:59 | 1 |
+------------------+----------------+------------------+------------------------+------------------+---------------------+----------------------+
[root@a2 ]# mysql -u root -p** -h a3 -e "select * from safeplan_new1.plan_approval where plan_approval_id=3247" +------------------+----------------+------------------+------------------------+------------------+---------------------+----------------------+
| plan_approval_id | plan_master_id | plan_location_id | plan_approval_location | plan_approval_by | plan_approved_on | plan_approval_status |
+------------------+----------------+------------------+------------------------+------------------+---------------------+----------------------+
| 3247 | 292 | 0 | 155 | 133 | 2013-07-03 16:13:59 | 1 |
+------------------+----------------+------------------+------------------------+------------------+---------------------+----------------------+
[root@a2 ]# mysql -u root -p** -h a4 -e "select * from safeplan_new1.plan_approval where plan_approval_id=3247" +------------------+----------------+------------------+------------------------+------------------+---------------------+----------------------+
| plan_approval_id | plan_master_id | plan_location_id | plan_approval_location | plan_approval_by | plan_approved_on | plan_approval_status |
+------------------+----------------+------------------+------------------------+------------------+---------------------+----------------------+
| 3247 | 292 | 0 | 155 | 133 | 2013-07-03 16:13:59 | 1 |
+------------------+----------------+------------------+------------------------+------------------+---------------------+----------------------+
Answer Answered by roberto spadim in this comment.
ok
run this query in each server
select to_unixtimestamp(plan_approved_on) from safeplan_new1.plan_approval where plan_approval_id=3247
(i don't know if the problem is the on update at slaves, i never tried it)
check if you have diferrent unixtimestamps for servers
after this we could put this history at jira (mariadb bug tracker - https://jira.mariadb.org), if we can't solve the error, i'm a mariadb user =]
propably they will ask about insert queires or binlog file
