#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
2014-09-18 08:15:03
otto_
good morning - are you serg online yet? I need some help understanding CMakeLists
2014-09-18 08:18:05
otto_
serg: hardening-wrapper got deprecated this summer in Debian and I am now trying to build without it: https://github.com/ottok/mariadb-10.0/commit/2497b4c0ca3e964712e5c3027833b3bdef3433c8
2014-09-18 08:20:28
otto_
serg: however I ran into some problems: 1) cmake/build_configurations/mysql_release.cmake line 138 triggers and stops build 2) if I force it to continue with flag -DIGNORE_AIO_CHECK=1 then build fails at /usr/share/cmake-2.8/Modules/TestBigEndian.cmake:44
2014-09-18 09:03:30
otto_
serg: ok, I figured it out now, reason was that the +all turned on PIE hardening the source does not seem to be able to handle. For now you can forget all I wrote above :)
2014-09-18 09:14:52
hartmut_
where in THD would i find an unique transaction identifier? doesn't have to be GTID, just something that changes when a new transaction is started explicitly or implicitly ...
2014-09-18 09:25:09
hartmut_
THD::transaction::all and ::stmt don't seem to be what i'm looking for ...?
2014-09-18 09:35:44
serg
hartmut: THD::transaction.xid_state.xid
2014-09-18 09:43:20
otto_
serg: another related question: on the page http://mentors.debian.net/package/mariadb-10.0 you can see complains on a few .so libs about https://lintian.debian.org/tags/hardening-no-fortify-functions.html - how do you interpret this? what is the correct thing to do about it?
2014-09-18 09:44:31
otto_
serg: here is the main Debian documentation about the topic: https://wiki.debian.org/Hardening
2014-09-18 09:45:14
hartmut_
serg: ah, didn't dig that deep yet ...
2014-09-18 09:45:15
serg
in 5.5 or 10.0 we don't fortify anything, debian does - it enables fortification in compiler flags
2014-09-18 09:45:48
serg
otto_: I've no idea why lintian complains about these two, I'd expect that everything is fortified or nothing
2014-09-18 09:46:17
serg
otto_: in 10.1 we fortify everything in all packages, from our CMakeLists.txt
2014-09-18 09:47:28
otto_
related, the issue earlier today was that hardening PIE caused the failures. However all other hardening mechanisms seemed to work, so I could maybe have the option 'DEB_BUILD_MAINT_OPTIONS = hardening=+all,-pie'
2014-09-18 09:48:29
otto_
but I feel a bit unsure, I need external confirmation to have the courage to run with these. Security build options and uncertainty is not a good combination..
2014-09-18 09:49:44
ekarlso
does 10.0.13 support xtradb yet or ?
2014-09-18 09:49:55
serg
otto_: that's why I didn't enable them in 5.5 or 10.0, but only in 10.1
2014-09-18 09:50:41
serg
ekarlso: yes, since 10.0 went GA in april.
2014-09-18 09:57:20
otto_
serg: ok, so I will not try to do more changes regarding hardening for now then
2014-09-18 10:42:41
frenda
Hi
2014-09-18 10:43:19
frenda
I'm going to install http://www.appropriateit.org/open-source-crowdfunding/#goteo
2014-09-18 10:43:33
frenda
Software Stack: Apache/php/mysql
2014-09-18 10:43:50
frenda
Can I install mariadb instead of mysql for Goteo
2014-09-18 10:43:53
frenda
?
2014-09-18 10:45:02
pomyk|w
yes
2014-09-18 10:46:12
frenda
Great.
2014-09-18 10:49:35
frenda
Is there something like phpmyadmin for mariadb?
2014-09-18 10:50:07
serg
phpmyadmin should work just fine
2014-09-18 10:50:25
frenda
found: https://mariadb.com/kb/en/mariadb/documentation/clients-and-utilities/graphical-clients/phpmyadmin/
2014-09-18 10:56:07
frenda
sould I still use mysql(d) as command for controlling MariaBD? (http://www.tecmint.com/install-lamp-in-arch-linux/)
2014-09-18 10:57:10
serg
yes
2014-09-18 15:24:02
n5
Hi, do we have RPMS of mariadb 10 to install in centos 7?
2014-09-18 15:24:40
n5
cant find any guide or repository
2014-09-18 15:28:24
pomyk|w
n5: not yet https://mariadb.atlassian.net/browse/MDEV-6433
2014-09-18 15:29:08
pomyk|w
n5: ius has it in testing repo: http://dl.iuscommunity.org/pub/ius/testing/CentOS/7/x86_64/
2014-09-18 15:31:54
n5
pomyk|w so its not advisable to use it in production yet?
2014-09-18 15:35:13
pomyk|w
i have no idea :)
2014-09-18 16:59:11
georg__
hartmut_, hartmut_ ping ?
2014-09-18 17:03:32
kolbe
knielsen: ping!
2014-09-18 17:29:31
cablop
hello
2014-09-18 17:31:33
cablop
i have this question... we are interested in MariaDB Galera Cluster... but we want to know if you icate ALL the databases in each member of the cluster, or if i can decide which databases are shared between which servers
2014-09-18 17:31:42
cablop
sigh
2014-09-18 17:31:47
cablop
let me rephrase
2014-09-18 17:39:31
dltl
the galera cluster looks pretty impressive - have you implemented it?
2014-09-18 17:40:02
cablop
we want to know if the members of a galera cluster should replicate ALL the database of all the other members, or if i can just decide who replicates what
2014-09-18 17:40:36
cablop
not yet... because we currently have multiple database servers... with different databases
2014-09-18 17:41:41
cablop
we want to setup a central server that replicates all the other members, but the other members don't replicate the databases that does not belong to them
2014-09-18 17:42:22
dltl
wouldn't you WANT that to happen?
2014-09-18 17:43:18
dltl
dltl sits back and hopes for some discussion
2014-09-18 17:49:12
grknight
cablop: that sounds like traditional replication . depends where the writes go
2014-09-18 17:50:04
cablop
mmm, let's say we two db servers: A and B with databases a1, a2, and b1, b2 and b3
2014-09-18 17:50:17
cablop
we want to start a new db server C
2014-09-18 17:50:24
cablop
and replicate this way
2014-09-18 17:50:59
cablop
A: a1, a2; B: b1, b2, b3; C: a1, a2, b1, b2, b3
2014-09-18 17:51:54
cablop
grknight: not exactly, because if someone modify something on C:b3 we want to see the changes in B:b3 too
2014-09-18 17:54:51
seattlegaucho
cablop, you need to start 2 replication streams in C
2014-09-18 17:55:07
seattlegaucho
1 to replicate from A and the 2nd one to replicate from B
2014-09-18 17:55:20
cablop
and how to replicate back?
2014-09-18 17:56:34
seattlegaucho
https://mariadb.com/kb/en/mariadb/documentation/replication-cluster-multi-master/replication/multi-source-replication/
2014-09-18 17:56:53
seattlegaucho
cablop, replicate from C to B and A too?
2014-09-18 17:57:39
cablop
yes
2014-09-18 17:58:57
cablop
oh, gosh... cluster does not support myisam tables... sigh
2014-09-18 17:59:46
grknight
you should be moving away from myisam. little reason to still use it
2014-09-18 18:00:02
seattlegaucho
the setup is the same, but you'll have to add 'filters' with 'replicate_ignore_db' (https://mariadb.com/kb/en/mariadb/documentation/replication-cluster-multi-master/replication/replication-and-binary-log-server-system-variables/#replicate_ignore_db)
2014-09-18 18:00:38
seattlegaucho
cablop, cluster replicates everything to every node, you can't filter
2014-09-18 18:01:38
cablop
ok... that's what we need to know
2014-09-18 18:01:55
cablop
then we are going to use replication for some weeks while we place all databases in the same server
2014-09-18 18:01:58
cablop
mmm
2014-09-18 18:02:00
cablop
wait
2014-09-18 18:02:33
cablop
if i add a server with no databases to a server with databases... will it start copying the databases from scratch?
2014-09-18 18:05:01
seattlegaucho
cablop, nope
2014-09-18 18:05:26
seattlegaucho
you need to preload to a given point in time and start replicating from that point in time
2014-09-18 18:36:32
kolbe
jplindst: ping!
2014-09-18 18:52:53
t0ms
Hi, we are having some weird issue with MariaDB 5.5... One of our client has around 1000 databases each with 100 tables running on MyISAM (...) and the whole server has really high %sys CPU usage
2014-09-18 18:52:59
t0ms
around 75-90%
2014-09-18 18:53:25
t0ms
the server is running CentOS 6.5 + ZFS (8k blocks, no compression) with enough memory and CPU power
2014-09-18 18:53:37
t0ms
any ideas why this might be happening ?
2014-09-18 18:54:40
tanj
t0ms: any iowait?
2014-09-18 18:54:47
t0ms
tanj: no
2014-09-18 18:54:57
tanj
t0ms: probably query optimization problem
2014-09-18 18:55:04
tanj
t0ms: how many QPS?
2014-09-18 18:56:03
tanj
t0ms: is it at least mysqld using those 75% CPU?
2014-09-18 18:56:26
t0ms
tanj: this is a dedicated box for mysql, it's the only service there
2014-09-18 18:56:52
dltl
1000 databases? wow
2014-09-18 18:57:05
t0ms
Queries per second avg: 2161.103
2014-09-18 18:57:48
t0ms
dltl: my bad, it's 4213 databases
2014-09-18 18:57:52
t0ms
..
2014-09-18 18:58:44
t0ms
each DB is having 148 tables
2014-09-18 18:59:04
t0ms
most of them are totally unused
2014-09-18 19:00:15
tanj
t0ms: please pastebin output of top -n1
2014-09-18 19:00:52
t0ms
http://pastebin.com/SarFHAmx
2014-09-18 19:00:53
t0ms
tanj: ^
2014-09-18 19:00:54
dltl
dltl hugs his 4 databases
2014-09-18 19:03:02
tanj
t0ms: that's lots of CPU for sure. Can you pastebin output of "SHOW FULL PROCESSLIST", "SHOW VARIABLES" and "SHOW GLOBAL STATUS" ?
2014-09-18 19:05:16
t0ms
tanj: http://pastebin.com/uE7QPDGE
2014-09-18 19:09:59
tanj
t0ms: many interesting things
2014-09-18 19:10:15
tanj
t0ms: your key buffer size is way too high, most of it is unused
2014-09-18 19:11:33
tanj
increase table_cache and table_definition_cache. You have 4200 databases so you need larger values
2014-09-18 19:12:13
tanj
wow! you have 2G query cache size :(
2014-09-18 19:12:17
tanj
this can be a huge problem
2014-09-18 19:12:22
tanj
reduce query cache size to 64M
2014-09-18 19:12:31
tanj
reduce key buffer size to 256M
2014-09-18 19:15:20
tanj
t0ms: are you there mate?
2014-09-18 19:17:26
dltl
dltl does all of tanj's suggestions in preparation for the future
2014-09-18 19:17:51
t0ms
tanj: yes, thanks, we are just looking at the variables with my colleague
2014-09-18 19:18:29
tanj
t0ms: you have two long queries running on information_schema. this can waste lots of resources
2014-09-18 19:18:45
dltl
dltl looks at the variables with the guy reflected in off the coffee mug
2014-09-18 19:18:50
tanj
t0ms: Query 1 and Query 13. The first one has been running for hours
2014-09-18 19:19:27
tanj
dltl: to check buffer and query cache use check in STATUS, key_blocks_unused and Qcache_free_memory
2014-09-18 19:19:40
tanj
if those two variables are high then you're doing it wrong :)
2014-09-18 19:19:51
tanj
and never never use a big query cache
2014-09-18 19:20:29
tanj
t0ms: you also have some slow queries, which are probably badly optimized
2014-09-18 19:21:22
t0ms
tanj: changed the query_cache to some normal number - 32M
2014-09-18 19:21:48
t0ms
also increased the table_cache to higher number
2014-09-18 19:21:59
t0ms
according to strace, the mysqld is stuck mostly in futex(2) call..
2014-09-18 19:22:22
tanj
t0ms: strace won't do much with mysqld - it is multithreaded
2014-09-18 19:22:34
tanj
t0ms: you need a DBA to look into the queries
2014-09-18 19:23:06
tanj
t0ms: some of them probably waste CPU because they are badly optimized. STATUS reports too much key scans or table scans (meanging = missing indexes or badly designed queries)
2014-09-18 19:23:13
t0ms
https://gist.github.com/snajpa/264353704fc30327dfca
2014-09-18 19:23:27
t0ms
tanj: this is an overview statistics from all current threads of MySQL from strace
2014-09-18 19:25:06
tanj
t0ms: I can't say whether this is normal or not. 1st of all I would try to get rid of queries 1 and 13 then repeat the operation
2014-09-18 19:26:01
t0ms
tanj: they are gone now
2014-09-18 19:26:16
tanj
t0ms: still have futexes/slow queries?
2014-09-18 19:26:29
t0ms
tanj: yes
2014-09-18 19:26:36
t0ms
no change
2014-09-18 19:26:53
t0ms
Cpu(s): 7.7%us, 89.9%sy, 0.0%ni, 2.4%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
2014-09-18 19:27:57
tanj
t0ms: as I said the cause of CPU use is probably bad query design
2014-09-18 19:28:11
ebergen
perf top?
2014-09-18 19:28:49
tanj
I think that futexes are mostly used by locking at the database engine level
2014-09-18 19:29:26
ebergen
do you have quickstack or gdb output?
2014-09-18 19:29:40
tanj
pmp maybe
2014-09-18 19:29:45
ebergen
pastebin i
2014-09-18 19:29:48
ebergen
*it
2014-09-18 19:30:37
tanj
t0ms: try the last example http://poormansprofiler.org/#!/bin/bashnsamples=1sleeptime=0pid=
2014-09-18 19:30:45
tanj
it will be more useful than strace
2014-09-18 19:40:16
t0ms
tanj: thanks, we are slightly modyfing those pormansprofiler's scripts :)
2014-09-18 19:40:29
t0ms
t0ms sighs... where is dtrace
2014-09-18 19:44:28
t0ms
tanj: https://gist.github.com/snajpa/d81de5659ce1767f51b9 is the output
2014-09-18 19:46:30
tanj
t0ms: not much to see, I would really look into query optimization at this point
2014-09-18 19:53:06
ebergen
it is all bottled up in myisam
2014-09-18 20:07:42
t0ms
tanj: thanks.. unfortunately no success
2014-09-18 20:08:58
tanj
t0ms: what do you mean by no success?
2014-09-18 20:09:09
t0ms
tanj: it's still slow
2014-09-18 20:09:27
t0ms
tanj: but, this might be a problem with ZFS... we will try to revert the version
2014-09-18 20:09:36
tanj
t0ms: I doubt that.
2014-09-18 20:09:53
tanj
t0ms: fs problem = iowait, no sys% on the mysqld process
2014-09-18 20:10:02
ebergen
what
2014-09-18 20:10:11
tanj
t0ms: It's still slow because you haven't optimized the queries.
2014-09-18 20:12:42
t0ms
tanj: I passed that to our client, thanks a lot for your time, i appreciate it!
2014-09-18 20:12:50
tanj
np
2014-09-18 20:16:45
tanj
ebergen: I meant that ZFS wouldn't cause the mysqld process to use more system CPU. Sorry. OTOH it would cause more iowait CPU if there were any issues
2014-09-18 20:17:14
tanj
system CPU is probably caused by mutex calls as far as my limited understanding goes
2014-09-18 21:06:04
inthl
is there any documentation on mariadb.com regarding server configuration? I can't find docs similar to the ones at dev.mysql.com
2014-09-18 21:14:37
tanj
inthl: yes, check mariadb.com/kb
2014-09-18 21:17:15
inthl
I was looking for https://mariadb.com/kb/en/mariadb/documentation/optimization-and-tuning/system-variables/server-system-variables/
2014-09-18 21:17:16
inthl
thanks :>
2014-09-18 23:32:52
kolbe
nirbhay_: ping
2014-09-18 23:33:03
nirbhay_
kolbe, yes
2014-09-18 23:33:36
kolbe
nirbhay_: did you have any luck repeating that INSERT/UPDATE galera crash?
2014-09-18 23:35:40
nirbhay_
kolbe, nope.. not with sync_wait values 0 and 1 so far.
2014-09-18 23:37:15
kolbe
ok
2014-09-18 23:37:27
kolbe
let me know if you want login information for these ec2 instances
2014-09-18 23:38:14
nirbhay_
sure.. are these test instances?
2014-09-18 23:39:26
kolbe
yeah
2014-09-18 23:39:41
kolbe
send me your public key and i'll stick it on 'em
2014-09-18 23:41:21
nirbhay_
kolbe, ack
2014-09-18 23:57:06
inthl
hm, regarding: ERROR 1227 (42000): Access denied; you need (at least one of) the SHOW DATABASES privilege(s) for this operation ...is it possible to assign that privilege for the user's databases only but not on a global scope, to see all dbs?
2014-09-18 23:58:10
ebergen
show databases is pretty boring if you only have access to one database
2014-09-18 23:59:42
mgriffin
information_schema!
2014-09-19 00:02:42
inthl
hm
2014-09-19 00:03:16
inthl
well I may have plenty of users and databases on one host, I'd like to prevent them from calling this, since mysql performs a directory-scan of the db-dir in that case
2014-09-19 00:16:30
Onlogn
Anyone read the Oracle news?
2014-09-19 00:17:56
Onlogn
I guess http://hurdcatz.com and http://hurdingcatz.com did...
2014-09-19 00:18:07
Onlogn
Although it goes to Bluehost once in a while...
2014-09-19 00:20:52
mgriffin
posted to #mysql eariler re: oracle http://www.cnbc.com/id/102014022
2014-09-19 00:21:50
dltl
oracle news -> no fixes to the pdf program ... no fixes to the mysql server ....
2014-09-19 00:23:00
dltl
oracle is what everyone blames google for trying to be
2014-09-19 00:23:30
kolbe
"fixes to the pdf program"?
2014-09-19 00:24:27
dltl
that acroread horror
2014-09-19 00:24:32
dltl
or whatever it is
2014-09-19 00:24:57
dltl
the one that has all the vulnerabilities they won't fix
2014-09-19 00:26:40
kolbe
no clue what you're talking about. i expect that something called acroread would be distributed by adobe intead of oracle, though?
2014-09-19 00:26:48
kolbe
or are you talking about some package in a repository for oracle linux?
2014-09-19 00:27:06
kolbe
i don't have any idea why i am even trying to talk to you about this, i have no idea what you're talking about and have no opinion about it :)
2014-09-19 00:29:40
dltl
hm? aren't adobe and oracle the same? who does flash?
2014-09-19 00:30:33
Jb_boin
nope
2014-09-19 00:30:39
dltl
which one gobbled up sun and destroyed it? nvm
2014-09-19 00:30:53
kolbe
what on earth...
2014-09-19 00:31:16
dltl
you try goggling sun to get an answer
2014-09-19 00:31:53
dltl
interesting article - and why are the sun and planets all in the same plane of rotation ... wth
2014-09-19 00:32:15
Jb_boin
you will end up burning your wings...
2014-09-19 00:32:54
dltl
dltl creates a 4th database to handle all the new info he found about the sun and planets ;)
2014-09-19 00:34:10
dltl
seriously - wasn't SUN originally call pacific network? i still have an early 68000 'sun/pacific' motherboard that still fires up
2014-09-19 00:35:08
kolbe
SUN = Stanford University Network
2014-09-19 00:35:47
dltl
unix all the way
2014-09-19 00:41:29
erkules
Jb_boin: If you are referring to Ikarus I like your answer :)
2014-09-19 00:43:34
mgriffin
this was before man invented glue and had to use wax to fly
2014-09-19 00:43:49
mgriffin
ah progress
2014-09-19 00:44:38
dltl
or invented staples
2014-09-19 00:47:39
erkules
:)
2014-09-19 00:49:54
dltl
jfyi - fpc has a very nice mysql library that works great with mariadb
2014-09-19 00:51:34
dltl
is there a list of 3rd party software that works well with mariadb? like libreoffice, phpmyadin and now fpc <-- my experiences that these work great
2014-09-19 02:47:22
erkules
ebergen: you are working @facebook right? https://www.facebook.com/notes/mysql-at-facebook/lessons-from-deploying-mysql-gtid-at-scale/10152252699590933 is some kind of irritating
2014-09-19 02:48:42
dltl
erkules: see if he can get you the password for misssexxy303030 secret photo cache ? ;)
2014-09-19 02:49:48
erkules
dltl: I know my password
2014-09-19 02:50:14
erkules
ebergen: it ignores the fact semisync sucks
2014-09-19 02:50:46
seekwill
Why does it suck
2014-09-19 02:50:56
erkules
because it is worthless
2014-09-19 02:51:56
erkules
so simply putting a failed master into the replication as a slave is likely to do not what you want