#maria Log v0.1

logs Catalogue

This page loads from the db, refresh as and when. It displays 24 hrs
For older data http://marialog.archivist.info/previous.php, logs go back to 2009-01-18
Some user stats
Table logs
date_time
user
message
2017-04-27 15:13:46
azathoth99
42
2017-04-27 15:16:44
rawi
azathoth99: thanks, 42 for key_buffer_size or myisam_sort_buffer_size?
2017-04-27 15:18:32
stemid
hi can anyone recognize this strange behavior I'm seeing? https://bpaste.net/show/5fd5e13e7107 cluster nodes randomly say they "timed out" and have to resync with an IST. this happens several times each day.
2017-04-27 15:19:31
stemid
I think semalmdledb04 stands out because it has a lot of different IPs and IDs in its timed out messages.
2017-04-27 15:19:47
stemid
even its own IP
2017-04-27 15:21:01
stemid
they're VMs and I've moved them to the same ESX host just to rule out networking issues.
2017-04-27 15:40:34
skape
Hi guys ! I'm new to docker and galera, so maybe I'm just missing something , I've started a cluster using the lastest container , so far is ok, but when I try to restart the first node in the env is the --wsrep-new-cluster still set so the container doesn't joint the cluster , How can I pass this parameter only once ???
2017-04-27 15:45:11
skape
can I put the parameter wsrep-new-cluster in the my.conf ? I know it doesn't make much sense but I can edit the file after the cluster is up
2017-04-27 16:40:29
shann
Hi
2017-04-27 16:41:27
shann
I create galera cluster with 3 nodes, i imported big database 8G on mysql-03, i show database on mysql-01 and mysql-02, but folder db size is 1,9M on mysql-03 13Go ? Why not synced, but for others database seem sync ok.
2017-04-27 16:48:52
mysql_ana
@shann, are you sure the nodes are sync in cluster ? run this command on node1 and check "show global status like 'wsrep_cluster_%';"
2017-04-27 16:50:53
shann
mysql_ana, seem info on mysql-03 and mysql-02, database imported in mysql-03
2017-04-27 16:51:14
shann
wsrep_cluster_status Primary on three nodes
2017-04-27 16:52:29
mysql_ana
@shann, do you have primary key on all your tables ?
2017-04-27 16:52:52
shann
on tables of new imported database ?
2017-04-27 16:53:04
mysql_ana
yes
2017-04-27 16:55:56
shann
i imported EZPublish i have 138 tables, i can use request to fast check them ?
2017-04-27 16:56:40
shann
also i note that another database seem not synced between node, ~200Mo on mysql-03 and other note ~100ko :(
2017-04-27 16:57:09
mysql_ana
@shann, you can check it, check in mysql-03 for all the tables if it has primary key
2017-04-27 16:59:42
mysql_ana
you may run this query to find the tables of interest
2017-04-27 16:59:46
mysql_ana
```SELECT t.table_schema, t.table_name FROM information_schema.tables t LEFT JOIN information_schema.table_constraints c ON (t.table_schema = c.table_schema AND t.table_name = c.table_name AND c.constraint_type IN ('PRIMARY KEY','UNIQUE')) WHERE t.table_schema NOT IN ('mysql','information_schema', 'performance_schema') AND t.engine NOT IN ('ARCHIVE','FEDERATED') AND c.table_name IS NULL;```
2017-04-27 16:59:47
shann
mysql_ana, i check table in case maskins, on mysql-03 show that primary key, on mysq-01 "Empty set"
2017-04-27 16:59:53
mysql_ana
SELECT t.table_schema, t.table_name FROM information_schema.tables t LEFT JOIN information_schema.table_constraints c ON (t.table_schema = c.table_schema AND t.table_name = c.table_name AND c.constraint_type IN ('PRIMARY KEY','UNIQUE')) WHERE t.table_schema NOT IN ('mysql','information_schema', 'performance_schema') AND t.engine NOT IN ('ARCHIVE','FEDERATED') AND c.table_name IS NULL;
2017-04-27 17:02:52
shann
i need execute query to mysql-03 or mysql-02, replace t for database name ?
2017-04-27 17:04:54
mysql_ana
@shann, no need to replace anything, run as it is. you need to run on mysql-03
2017-04-27 17:06:37
shann
mysql_ana, ok, i launch command, 22 rows on 6 databases
2017-04-27 17:08:13
shann
Table './<database_name>/<table>' is marked as crashed and last (automatic?) repair failed :(
2017-04-27 17:09:06
mysql_ana
@shann
2017-04-27 17:09:16
mysql_ana
those 22 rows are the tables without primary key
2017-04-27 17:09:25
mysql_ana
fix it and run, galera needs primary key
2017-04-27 17:09:32
mysql_ana
i got to go now, sorry cnat continue
2017-04-27 17:09:48
mysql_ana
good luck
2017-04-27 17:10:52
shann
mysql_ana thanks lot for way to fix issue i check them :)
2017-04-27 17:24:48
serg
svoj?
2017-04-27 17:26:58
svoj
serg: ?
2017-04-27 18:17:47
shann
i create index on table where is missing, for now two databases don't have all tables with index, but other databases not synced, need all databases have indexed tables from sync work ?
2017-04-27 18:57:01
shann
mysql_ana, i clean all databases, force mysql-01 and mysql-02 to resync, now it's work perfectly, all node synced, thanks again for your help.
2017-04-28 09:03:57
markusjm
svoj: I was listening in on your talk about atomic operations at the developer unconference and I have a quick question
2017-04-28 09:06:12
markusjm
can a __sync_synchronize() be used to replace a call to a __atomic_store()?
2017-04-28 09:09:57
svoj
markusjm: it can't according to standard
2017-04-28 09:10:13
markusjm
hmm, OK
2017-04-28 09:10:59
svoj
also __sync and __atomic are different builtins, I wouldn't mix those
2017-04-28 09:12:22
markusjm
we're mainly facing a problem with CentOS 6 which doesn't support __atomic builtins
2017-04-28 09:12:35
svoj
strictly speaking __sync_synchronize() issues memory barrier, while __atomic_store() guarantees atomicity
2017-04-28 09:12:36
markusjm
and I'm looking for a suitable way to implement an atomic store/load
2017-04-28 09:13:00
svoj
why don't you like __atomic_store and __atomic_load?
2017-04-28 09:13:09
markusjm
CentOS 6 doesn't have them :)
2017-04-28 09:13:15
markusjm
I do like them :D
2017-04-28 09:13:26
svoj
did you check how server implements them with sync builtins?
2017-04-28 09:13:43
markusjm
oh yeah, I didn't realize to check ther
2017-04-28 09:13:51
markusjm
any pointers on where to look at?
2017-04-28 09:14:17
svoj
#define make_atomic_load_body(S) ret= __sync_fetch_and_or(a, 0);
2017-04-28 09:14:18
svoj
#define make_atomic_store_body(S) __sync_lock_test_and_set(a, v);
2017-04-28 09:14:32
svoj
10.3/include/atomic/gcc_builtins.h
2017-04-28 09:14:56
svoj
ugly, but it works
2017-04-28 09:15:38
markusjm
I get the fetch_and_or but what does the test_and_set do?
2017-04-28 09:15:47
svoj
and strictly speaking store is broken: it issues acquire instead of seq_cst
2017-04-28 09:15:57
svoj
https://gcc.gnu.org/onlinedocs/gcc/_005f_005fsync-Builtins.html
2017-04-28 09:16:30
svoj
writes new value, returns old
2017-04-28 09:16:31
markusjm
oh yeah, the naming was very confusing for those :D
2017-04-28 09:16:38
markusjm
so it's just atomic exchange
2017-04-28 09:17:04
svoj
right, not perfect, but still better than nothing
2017-04-28 09:17:07
markusjm
true
2017-04-28 09:18:01
markusjm
assuming we're on an x86_64 processor, would a __sync_synchronize() with a normal store work like __atomic_store()?
2017-04-28 09:18:22
markusjm
any clues? :D
2017-04-28 09:18:56
markusjm
I better go and look at the compiled instructions
2017-04-28 09:19:05
markusjm
I might even learn a thing or two :)
2017-04-28 09:19:29
svoj
I'd say I have no examples that wouldn't work. But keep in mind that compiler can do tricks even on x86_64.
2017-04-28 09:20:14
svoj
so it's not only about platform, also about compiler optimizations
2017-04-28 09:34:31
markusjm
hmm, true
2017-04-28 09:34:50
markusjm
but I think I'll go with the "if it works in MariaDB server, I'll trust it works" methodology of coding :)
2017-04-28 09:35:05
markusjm
and use __sync_lock_test_and_set for atomic store
2017-04-28 10:46:33
serg
grknight: generic gentoo question. after a bug (in gentoo bugzilla) is closed with a reference to a commit, how long is it till "emerge --sync" will fetch it?
2017-04-28 11:03:30
lapinrigolo
hello. I'm trying to add DBUG_PRINT statements to my MariaDB debug build but nothing ever comes up on the .trace file
2017-04-28 11:03:53
lapinrigolo
I know the server's been built in debug mode and I can see the .trace file being updated
2017-04-28 11:04:12
lapinrigolo
it's just the DBUG_PRINT statements that seem to be ignored
2017-04-28 11:14:24
serg
lapinrigolo: it needs to be enabled with the --debug (or -#) option. like -#d or -#d:o,dbug.log
2017-04-28 11:14:58
serg
or selectively your DBUG_PRINT keyword. see dbug/user.ps or dbug/user.t
2017-04-28 11:16:52
lapinrigolo
serg I start the server with the --debug flag
2017-04-28 11:17:12
lapinrigolo
and I add DBUG_PRINT("info", ("aaa")); statements, but these don't seem to be printed
2017-04-28 11:17:36
lapinrigolo
I can see other DBUG_PRINT statements on the trace file, only mine seem to be missing
2017-04-28 13:25:55
grknight
serg: 30-60 minutes until the commit is sent to all of the rsync mirrors
2017-04-28 14:17:14
darkfrog
I'm on Ubuntu 17.04 and when I install mariadb-server-10.1 `mysql -u root` denies me access....anyone know why?
2017-04-28 14:21:22
LOBINIOS
ciao
2017-04-28 14:21:29
LOBINIOS
!list
2017-04-28 14:25:25
grknight
darkfrog: some distros are forcing the unix user root to connect as 'root'@'localhost' (socket), no other user will have access
2017-04-28 14:26:35
grknight
they are doing this for their maintenance scripts. it is possible to disable this feature, but said scripts will fail
2017-04-28 14:27:31
grknight
you can also create a full priviledge user in addition to the default 'root'@'localhost'
2017-04-28 14:29:16
darkfrog
grknight: I used "sudo mysql -u root" to log in and then completely re-created the root user...that fixed it. Thanks. :)