#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
2015-03-26 18:31:41
bytee
svoj: nice blog !
2015-03-26 18:32:24
svoj
bytee: thanks!
2015-03-26 20:52:21
bytee
sigor: howdy
2015-03-26 21:50:49
bytee
lawrin: hey, where is odbc linux ?
2015-03-27 01:46:55
mgriffin
what is a good starting gcache.size where there is plenty of storage and lots of writes? 2x innodb_log_file_size?
2015-03-27 01:47:46
mgriffin
If so and I set innodb_log_file_size=256M and gcache.size=512M what do I set gcache.page_size to? 10G?
2015-03-27 01:56:07
mgriffin
The docs for this are kinda confusing IMO
2015-03-27 02:26:34
danblack
mgriffin: i generally got for about 1G. enough to be read quicky and enough there to recover a node if its offline for a bit without a stst
2015-03-27 02:26:37
danblack
sst
2015-03-27 02:27:17
ebergen
what is a gcache?
2015-03-27 02:27:30
danblack
galera cache
2015-03-27 02:27:48
mgriffin
danblack: i went with 1G, as well. thanks for the advice. now i am struggling with ssl.
2015-03-27 02:36:20
danblack
i haven't battled galera ssl yet. good luck. multicast has finiky enough.
2015-03-27 02:50:48
mgriffin
danblack: i think i was just too lazy creating the certs
2015-03-27 02:55:50
danblack
ssl tends to be a bit unforgiving there
2015-03-27 03:03:46
mgriffin
i thought i had to just use different CN for the ca and just <enter> for *everything* else, which might not be quite right
2015-03-27 03:03:57
mgriffin
following https://dev.mysql.com/doc/refman/5.0/en/creating-ssl-certs.html
2015-03-27 03:06:02
mgriffin
this was sufficient for the mysql client btw
2015-03-27 03:06:20
mgriffin
but testing with s_client shows some error which leads down bad cert path
2015-03-27 03:06:44
mgriffin
Client: SSL:Cipher in use is DHE-RSA-AES256-SHA
2015-03-27 03:07:02
mgriffin
Galera/s_client: Verify return code: 21 (unable to verify the first certificate)
2015-03-27 03:07:22
mgriffin
SSL routines:SSL3_GET_CLIENT_CERTIFICATE:no certificate returned
2015-03-27 03:07:25
mgriffin
*sigh*
2015-03-27 03:09:53
mgriffin
via http://stackoverflow.com/questions/27871231/verifying-ssl-client-authenticity-fails-due-to-ssl3-get-client-certificateno-ce
2015-03-27 03:39:31
mgriffin
using a unique CN for all three fixed it
2015-03-27 03:40:52
mgriffin
ah perhaps not, ignore me.
2015-03-27 03:42:44
danblack
no, but we'll wait for the right solution :-)
2015-03-27 03:48:06
mgriffin
I did learn that rhel does something funny with sudo:
2015-03-27 03:48:33
mgriffin
sudo: sorry, you must have a tty to run sudo
2015-03-27 03:48:50
mgriffin
# cat /etc/sudoers.d/mysql
2015-03-27 03:48:50
mgriffin
mysql ALL = (root) NOPASSWD: /sbin/iptables
2015-03-27 03:48:51
mgriffin
Defaults:mysql !requiretty
2015-03-27 03:48:59
mgriffin
(fixed)
2015-03-27 03:50:00
danblack
how is mysql doing sudo? for xtrabackup sst? does it need it?
2015-03-27 03:51:52
mgriffin
danblack: http://www.fromdual.com/sites/default/files/block_galera_node.shx
2015-03-27 03:53:39
danblack
ah. interesting script use. why people insist on using iptables --list output to parse instead of iptables-save i've no idea.
2015-03-27 04:42:02
mgriffin
danblack: fixed *sigh*
2015-03-27 04:44:55
mgriffin
Docs bug in Galera, Maria, and Percona, none mention ssl_ca which solved it (or using a self with no ca)
2015-03-27 04:45:13
mgriffin
socket.ssl_ca=/etc/mysql-ssl/ca.pem = magic
2015-03-27 04:46:12
thumbs
almost warrants a blog post!
2015-03-27 04:46:29
mgriffin
thumbs: people been complaining for years: http://www.percona.com/blog/2013/05/03/percona-xtradb-cluster-for-mysql-and-encrypted-galera-replication/
2015-03-27 04:46:38
mgriffin
but not filing a bug report ;)
2015-03-27 04:47:34
mgriffin
i, for one, welcome our new LibreSSL overlords
2015-03-27 04:48:37
mgriffin
http://www.libressl.org/goals.html
2015-03-27 04:53:52
danblack
mgriffin: lucky for you http://galeracluster.com/documentation-webpages/galeraparameters.html has an edit on github link.
2015-03-27 04:55:50
danblack
mgriffin: should see the nginx-dev list trying to keep compatible with libressl while they rip out features and they seem to refuse to give a #define of libressl so compatible code can be written. I may have got it slightly wrong there but it all looks quite tense
2015-03-27 05:07:02
mgriffin
danblack: https://code.launchpad.net/~jheiss/galera/galera/+merge/194296
2015-03-27 05:10:14
danblack
https://github.com/codership/galera/blob/4c23a5bc11d8de9ad215bad3f268dd1348d3d9c4/galerautils/src/gu_asio.hpp#L55
2015-03-27 05:11:37
mgriffin
danblack: oh you found the docs (the source code)
2015-03-27 05:12:36
danblack
oh, they like pulling in doc fixed. its code they have troubles with https://github.com/codership/galera/pull/109
2015-03-27 05:15:12
danblack
https://github.com/codership/galera/issues/256 :-) couldn't be bothered writing it for them
2015-03-27 05:19:23
mgriffin
danblack: thanks!
2015-03-27 05:46:08
mgriffin
https://github.com/codership/galera/blob/3.x/garb/files/garb.sh#L145 also seems to be a bug :(
2015-03-27 05:46:55
mgriffin
runuser nobody -s /bin/sh -c '/usr/bin/garbd -d -a gcomm://1.2.3.4:4567 -g foo -o socket.ssl_cipher=AES128-SHA;socket.ssl_ca=/etc/mysql-ssl/ca.pem;socket.ssl_cert=/etc/mysql-ssl/server-cert.pem;socket.ssl_key=/etc/mysql-ssl/server-key.pem;'
2015-03-27 05:47:27
mgriffin
[ -n "$GALERA_OPTIONS" ] && OPTIONS="$OPTIONS -o '$GALERA_OPTIONS'"
2015-03-27 05:47:40
mgriffin
seems to work but I often get smacked down on my bash fu
2015-03-27 05:48:58
mgriffin
https://github.com/codership/galera/issues/186
2015-03-27 05:49:35
danblack
yes. ; need escaping again perhaps.
2015-03-27 05:55:22
mgriffin
updated the issue just now, feel free to improve
2015-03-27 05:56:39
danblack
i think i've done my share of bug reports for today. hopeing to get though the day without writing a new one.
2015-03-27 08:44:54
umarzuki
mariadb-galera-server : Depends: mariadb-galera-server-10.0 (= 10.0.17+maria-1~trusty) but it is not going to be installed
2015-03-27 08:45:15
umarzuki
I got the above when trying to install galera-mariadb
2015-03-27 08:45:31
tanj
umarzuki: what is the apt line you're using exactly?
2015-03-27 08:46:00
umarzuki
tanj: sudo LC_ALL=en_US.utf8 DEBIAN_FRONTEND=noninteractive apt-get -o Dpkg::Options::='--force-confnew' -y install mariadb-galera-server galera
2015-03-27 08:46:27
umarzuki
I'm migrating/upgrading from 5.5
2015-03-27 08:46:48
tanj
umarzuki: there's an issue happening if the universe repo is enabled. can you check if this is the case?
2015-03-27 08:47:15
tanj
you can try: sudo apt-get install mariadb-server-core-10.0
2015-03-27 08:47:27
tanj
this should pull the whole server foo
2015-03-27 08:48:07
tanj
or you could remove the 5.5 packages altogether and install from scratch
2015-03-27 08:48:18
tanj
there's some conflict between our repo and the ubuntu repo
2015-03-27 08:51:02
umarzuki
tanj: I already remove all packages named maria
2015-03-27 08:51:19
umarzuki
I also disabled universe
2015-03-27 08:51:40
umarzuki
does "deinstall" status means it is not completely removed?
2015-03-27 08:55:51
umarzuki
tried purging them but same problem
2015-03-27 09:12:53
tanj
umarzuki: can you show the output of dpkg -l | grep maria
2015-03-27 09:13:41
umarzuki
tanj: rc mysql-common 5.5.41+maria-1~trusty all MariaDB database common files (e.g. /etc/mysql/my.cnf)
2015-03-27 09:15:22
tanj
umarzuki: please try the following: aptitude install mariadb-server-core-10.0 and tell me if there is any error messages.
2015-03-27 09:16:17
umarzuki
tanj: no error up to [Y/n]
2015-03-27 09:16:37
umarzuki
i did not press enter yet
2015-03-27 09:17:00
tanj
what's the offered solution
2015-03-27 09:17:29
umarzuki
libmariadbclient18{a} libmysqlclient18{a} mariadb-common{a} mariadb-server-core-10.0 mysql-common{a}
2015-03-27 09:25:20
loomsen
hi folks, I noticed some strange behavior. I updated the mariadb-galera-server on one of our nodes, but upon restart, it fails to join the cluster (trying rsync receive address to be 127.0.0.1 instead of the LAN IP). When I set wsrep_receive_address rsync works, but the node still shows up as 127.0.0.1 in wsrep status
2015-03-27 09:26:35
loomsen
| wsrep_incoming_addresses | 172.17.50.46:3306,172.17.50.48:3306,127.0.0.1:3306 | # this is the output from node 172.17.50.46, the node in question is 172.17.50.47, but as you can see, it shows up as 127.0.0.1
2015-03-27 09:26:59
loomsen
what's wrong here please
2015-03-27 09:27:10
loomsen
had this problem both for 5.5 and 10.0
2015-03-27 09:29:19
vicentiu
all
2015-03-27 09:30:34
umarzuki
tanj: now its "mariadb-galera-server : Depends: mariadb-galera-server-10.0 (= 10.0.17+maria-1~precise) but it is not going to be installed" after I changed trusty tp precise for mariadb repo line
2015-03-27 09:36:02
loomsen
however, the replication is functional
2015-03-27 09:36:10
loomsen
the address is just showing up incorrectly
2015-03-27 09:37:36
loomsen
if I set the bind_address to the LAN IP, it shows up correctly
2015-03-27 09:40:02
loomsen
... ok it seems to be a buggy behavior with venet interfaces from openvz, which get both, loopback and LAN IP set
2015-03-27 09:46:29
chick
Hi I have dropped a proposal on melange for the GSoC GIS enhancement, however I wish to work on this related bug : MDEV-7540 while waiting for google to make a decision.
2015-03-27 09:46:29
chick
I build mariadb from a github clone but when I run the query described: SELECT * FROM INFORMATION_SCHEMA.SPATIAL_REF_SYS; I get an error not a result set as reported: http://pastebin.com/YWHGCbpK
2015-03-27 09:53:45
danblack
chick: Affects Version/s:
2015-03-27 09:53:47
danblack
10.1.2
2015-03-27 09:54:02
danblack
5.5.41 isn't late enough
2015-03-27 09:57:44
chick
danblack, ^^^
2015-03-27 10:05:54
umarzuki
tanj: it was settled, just installed manually and now running 10.0
2015-03-27 10:42:32
Nebraskka
Hey there! Let's say 1 of 3 Galera nodes being in SYNC state. What would happened if application still use it to SELECT and INSERT/UPDATE?
2015-03-27 10:42:54
Nebraskka
it would proxy them to donor node, or would mess up with data?
2015-03-27 11:05:49
danblack
chick: from the mdev. make sense. you need version 10.1 to try to fix this bug.
2015-03-27 11:06:31
danblack
Nebraskka: it will queue up changes for the other nodes. the node has to be primary however. try it and see if you want.
2015-03-27 11:08:39
Nebraskka
danblack, so syncing galera node is not usable right after it started since crash?
2015-03-27 11:08:41
chick
Yes danblack! I didnt look at the Affects Version/s: field of the bug report. Im switching to 10.1.2. Thankx
2015-03-27 11:09:33
danblack
chick: happy bug hunting. if you find any more along the don't forget to lodge them.
2015-03-27 11:09:54
Nebraskka
danblack, i'm thinking of makind 3 nodes "app->galera" in 3 DC, and thinking of it's cons
2015-03-27 11:10:53
Nebraskka
trying to find out what would app of crashed galera node see while it's galera instance being in SYNC state
2015-03-27 11:11:02
Nebraskka
outdated info?
2015-03-27 11:14:37
danblack
Nebraskka: a synced state has the latest info and therefore it isnt' a crashed node.
2015-03-27 11:15:11
danblack
there is a finite read delay so you can read out of date information.
2015-03-27 11:16:49
danblack
how often do each of the nodes touch the same data? is it important if they do or if a read delay happens. Write conflicts are detected and deadlock exceptions are thrown on commit. make sure you app handles those
2015-03-27 11:19:57
Nebraskka
danblack, so if galera node still transfer data from it's donor, SELECT/INSERT to this node would apply to this node and not the donor? it would result in some read delay, limited by timeout that would show up outdated data in the end if it's not still synchronized?
2015-03-27 11:20:30
danblack
a node won't talk to an application if it isn't in SYNC
2015-03-27 11:20:41
Nebraskka
now i see
2015-03-27 11:21:02
Nebraskka
:) many thanks
2015-03-27 11:21:13
danblack
good luck.
2015-03-27 14:35:44
Mariane
Hi I downloaded mariadb source code from github compiled it successfully but when i try to start in from the BUILD/client directory like this: ./mysql -u root i get this error: Can't connect to local MySQL server through socket '/var/run/mysqld/mysqld.sock'
2015-03-27 14:36:48
grknight
Mariane: mysql is the client. mysqld is the server
2015-03-27 14:37:50
Mariane
grknight, How do I start the server then? my build files are written to ~/server/BUILD/client
2015-03-27 14:38:06
grknight
the mysql command line client can be used with any mysql or mariadb server
2015-03-27 14:38:54
grknight
to start the server, you must first run mysql_install_db to install the system database, then start mysqld either directly or with the mysqld_safe script
2015-03-27 14:39:33
thumbs
if you run mysql_install_db manually, make sure you run mysql_secure_installation --user=mysql
2015-03-27 14:40:41
Mariane
from what directory can i run mysql_install_db?
2015-03-27 14:41:06
thumbs
Mariane: any.
2015-03-27 14:43:34
Mariane
thumbs, so after running make I should run mysql_install_db?
2015-03-27 14:49:00
Mamykin
Hello guys, can someone try to explain this? :) Aborted connection 540 to db: 'unconnected' user: 'replicator' host: 'ballouportalen-sql02' (Unknown error)
2015-03-27 14:55:37
topkatz
hello, Im using centos& and my mariadb will some times not start when issueing systemctl start mariadb
2015-03-27 14:55:49
topkatz
when this hapens it times out and I get no log message
2015-03-27 14:55:59
topkatz
it just says time out and does not start
2015-03-27 15:01:55
tanj
topkatz: systemctl shouldn't time out.
2015-03-27 15:02:18
tanj
topkatz: try journalctl -xe or journalctl -u maria to see the errors
2015-03-27 15:02:24
topkatz
ok
2015-03-27 15:02:31
topkatz
thanks
2015-03-27 15:02:57
topkatz
could it be a large innodb_buffer_pool size?
2015-03-27 15:03:45
topkatz
it can start and stop it a couple of times and then it just times out
2015-03-27 15:04:43
thumbs
how large?
2015-03-27 15:04:57
topkatz
my vm has 32gb and I gave it 10g
2015-03-27 15:07:58
topkatz
I droped it down to 1g
2015-03-27 15:08:26
topkatz
its letting me issue start|stops now. However I find it works better after a reboot, which I just did
2015-03-27 15:08:54
topkatz
the vm is running more then just mysql. Maybe I was to agresive and running out of memory?
2015-03-27 15:09:02
thumbs
topkatz: maybe.
2015-03-27 15:09:09
thumbs
topkatz: what else do you run on it?
2015-03-27 15:09:29
topkatz
apache, ZRM ZMC, jasper reports server
2015-03-27 15:09:35
topkatz
its sort of a self contained app
2015-03-27 15:10:13
topkatz
LAMP stack with backup and report server
2015-03-27 15:10:18
thumbs
topkatz: how much RAM is each component using?
2015-03-27 15:10:34
topkatz
not sure exactly
2015-03-27 15:10:40
thumbs
topkatz: also, if you use mod_php with apache httpd and prefork, don't.
2015-03-27 15:10:58
thumbs
topkatz: use the event mpm with httpd, and configure php-fpm pools
2015-03-27 15:12:41
topkatz
looks like I have both prefork and mod_php5 loaded
2015-03-27 15:13:09
thumbs
topkatz: that's terrible.
2015-03-27 15:13:33
topkatz
its default out of the box from centos
2015-03-27 15:14:03
thumbs
topkatz: still terrible.
2015-03-27 15:40:42
wp92
guys, i have a little problem :) i am trying to upgrade a platform using replication and 5.0.45. It is heavy stored procedure used, and in the binlog, every update/insert is overheaded by name_const(xxx,latin1('yyy')). if i upgrade a slave to 5.1, replication breaks with illegal mix of collation (tables are latin1_general_ci). In 5.0.72, my problem is fixed by adding the collation as well in the name_const() function, so replication
2015-03-27 15:40:42
wp92
doesn't break. I upgraded a slave to 5.0.96, and collation is already enforced in this version. Is there a way to disable this enforcing (same way as 5.0.45) ?