#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
2014-10-31 13:36:34
knielsen
serg, elenst: I tried to implement something in Buildbot with categories, as a follow-up from what we discussed a few days ago
2014-10-31 13:36:55
knielsen
I put all builders into one of just two categories: "main" and "experimental".
2014-10-31 13:37:05
elenst
knielsen: did you find out that we can actually hide 'experimental' by default?
2014-10-31 13:37:18
knielsen
The "main" builders should be kept 100% green at all times. The experimental can be used at will
2014-10-31 13:37:58
knielsen
elenst: yes, I also implemented a "template" for the grid.html page. It will redirect by default to a page showing only "main". And it has links to show all / just experimental
2014-10-31 13:38:09
elenst
ah wonderful
2014-10-31 13:38:21
knielsen
Let me know if it works out, or if something needs to be tweaked
2014-10-31 13:38:38
elenst
okay, lets watch how it goes a bit and see
2014-10-31 13:38:46
knielsen
elenst: I kept the categories for rqg* tests I believe, but feel free to change them if needed (or ask me to change them)
2014-10-31 13:39:02
elenst
yes, i'll change if needed, no problem
2014-10-31 13:39:36
knielsen
For now I put the "connector" type builders into "experimental" (mostly to not clutter the main trees with empty cells). Maybe we can add a "connector" category, if needed/useful
2014-10-31 13:40:00
elenst
yeah we can ask georg how he prefers it
2014-10-31 13:40:56
elenst
elenst 's long term delusion becomes a reality... i have no idea why i was sure that new and experimental were hidden by default -- that's what happens when you don't use the general links
2014-10-31 13:41:08
knielsen
hehe
2014-10-31 13:41:40
knielsen
elenst: It's actually pretty neat that there is a template system, it is quite easy to customise the builbot pages. We can probably use that for other neat things
2014-10-31 13:42:04
knielsen
knielsen kind of knew there was something along those lines, but didn't try it before
2014-10-31 13:42:14
elenst
right
2014-10-31 13:46:09
elenst
knielsen: we can't have win*-packages experimental
2014-10-31 13:46:17
elenst
they build actual release packages
2014-10-31 13:46:36
knielsen
elenst: well, then someone needs to fix them so they run green at all times with no false positives
2014-10-31 13:46:41
knielsen
then they can be moved to main
2014-10-31 13:47:37
elenst
but that's a catch 22 -- if they stay in experimental, nobody will bother to fix them and only few will even see them. They look pretty green to me on 5.5 btw, lets see 10.0
2014-10-31 13:48:14
knielsen
elenst: Well, it wasn't me who asked to have builders in the tree that were allowed to have false positives
2014-10-31 13:48:26
elenst
green in 10.0 too
2014-10-31 13:49:12
elenst
everything has false positives, but these two actually seem pretty stable to me -- what made you choose them as bad guys?
2014-10-31 13:49:44
knielsen
the win builders were never stable, random "failed bzr" and so on.
2014-10-31 13:50:09
knielsen
But it was mostly because I am not willing to keep those free of false positives, it's more than enough to do the unix ones
2014-10-31 13:50:22
knielsen
but sure, the more in "main", the better
2014-10-31 13:51:19
elenst
knielsen: anyway, we can't hide them, they do build mandatory release packages. And we must have windows presence in the main, because there are windows-specific regressions all the time. I agree that main must be kept green, so lets add them to main and next time they start failing make it a critical task to fix them?
2014-10-31 13:51:53
knielsen
elenst: ok
2014-10-31 13:52:09
knielsen
elenst: I'll add win32-packages and winx64-packages
2014-10-31 13:52:14
elenst
ok
2014-10-31 13:52:17
elenst
thanks
2014-10-31 13:53:36
elenst
knielsen: but you do realize that it will happen to every builder -- everything is bound to fail from time to time due to network problems and what not -- we can't exclude builders if it only happens once in a while, and goes away after the build is forced to restart
2014-10-31 13:54:32
elenst
if we kick off a builder due to a sporadic bzr problem, we'll end up with no builders whatsoever
2014-10-31 13:54:45
knielsen
I'm not trying to end up with no bulders
2014-10-31 13:55:02
elenst
well okay, i take it back, not all builders use bzr -- but you know what i mean
2014-10-31 13:55:26
knielsen
But whenever there is a problem, yes, we can restart the build, but we also need to do something to prevent the failure from happening again
2014-10-31 13:55:57
knielsen
When network issues becomes the main problem, I'll try to implement auto retry of bzr / git checkout in case it fails
2014-10-31 13:56:22
elenst
disk space is our main problem
2014-10-31 13:56:27
elenst
we can start from that
2014-10-31 13:56:29
knielsen
You're right of course that we can never get 100%, but we can get a lot closer than what we have now
2014-10-31 13:56:36
elenst
right
2014-10-31 13:56:40
knielsen
yes. I already did several fixes in that regard
2014-10-31 13:56:46
knielsen
(I mean in regard to disk space)
2014-10-31 13:56:52
elenst
yes, i got it, great
2014-10-31 13:57:40
elenst
knielsen: are you also going to make templates for other views (tgrid and whatever else is there)?
2014-10-31 13:57:49
elenst
i personally don't use them, but maybe others do
2014-10-31 13:57:55
elenst
elenst only uses grids
2014-10-31 13:58:01
knielsen
I can do it if somebody wants it
2014-10-31 13:58:12
knielsen
But for now I wasn't planning to, I also use only grid...
2014-10-31 13:58:22
elenst
okay
2014-10-31 14:00:40
knielsen
elenst: Maybe there are more builders that should be moved from "experimental" to "main". I mostly did the "experimental" because you and serg said you wanted to have some builders that you knew could not be expetected to never fail normally, but that were still useful to have in buildbot to use the whole build infrastructure
2014-10-31 14:01:40
elenst
knielsen: yes, it's certainly true for my win-rqg-se builder, by its nature
2014-10-31 14:01:57
elenst
other rqg builders -- i'll take a look, but I agree they're not stable enough to have them in main
2014-10-31 14:02:14
elenst
and nobody apart from me will investigate their faiures anyway
2014-10-31 14:02:44
knielsen
anyway, as you said, we can try this for a while and see how it goes
2014-10-31 14:02:51
elenst
right
2014-10-31 14:03:50
knielsen
For now, I think I fixed all the non-green main builders in 5.5 (except a recent valgrind that spetrunia2 promised to fix), we'll see if it holds after build of latest revision. If so, I'll start working on 10.0...
2014-10-31 14:04:12
elenst
knielsen: i'm surprised you've put centos7 in the 'main' category
2014-10-31 14:04:37
elenst
afair the whole discussion started from it -- it's as experimental (== new) as it gets
2014-10-31 14:05:07
knielsen
Ok, I can move it to experimental
2014-10-31 14:05:41
elenst
well, it's up to you, but i don't expect it to become green in the nearest future
2014-10-31 14:05:45
elenst
like, in days
2014-10-31 14:11:47
elenst
knielsen: i'm also not sure about *-dan-* builders. I don't really know what they are, but the debug one apparently has been down for ages, and the release one always seemed very unstable to me. Now I see that it's decently green in 5.5, so not sure about its status, maybe you've fixed it?
2014-10-31 14:12:44
knielsen
yeah, you're probably right about bld-dan-debug, I suppose...
2014-10-31 14:13:55
knielsen
bld-dan-release is rather important, as it builds without performance schema, I don't think we have other builders that test this...
2014-10-31 16:37:00
dcmwai
Hello there would like to check about the mariadb-cluster function
2014-10-31 16:37:17
dcmwai
I saw guide that it required 3 node mininal..
2014-10-31 16:37:25
dcmwai
what can I do if I've only 2?
2014-10-31 17:04:00
Crytponix_
salle: Hello
2014-10-31 17:08:03
Crytponix_
salle: If you recall, you had sent me a few examples for hardware acceleration, where you mentioned the time it takes for each step to execute in an SQL query
2014-10-31 17:08:24
Crytponix_
salle: This was the example you sent me: pastie.org/9669982#19
2014-10-31 17:08:48
Crytponix_
salle: I was wondering, do you remember how many records there were in the table, of which 25 rows were returned?
2014-10-31 17:09:12
salle
Crytponix_: It wasn't about hardware acceleratioin and I don't remember how many rows
2014-10-31 17:09:12
Naktibalda
25 is a number of rows in SHOW PROFILE output
2014-10-31 17:09:28
salle
Crytponix_: Try it with various queries
2014-10-31 17:09:50
Crytponix_
salle: Ah ok, Thanks
2014-10-31 17:09:59
Crytponix_
Naktibalda: Thanks for clarification
2014-10-31 17:13:36
salle
Crytponix_: Probably this table: http://pastie.org/9688139
2014-10-31 19:14:17
georg_
Pierre around?
2014-10-31 19:49:07
Crytponix_
salle: Thanks. Well, there are 512,000 records. Taking the "Grouping" into account, it has be processed fast I would say. I assume you have a good CPU or SSD Disks?
2014-10-31 19:50:42
salle
Crytponix_: desktop PC
2014-10-31 19:51:10
salle
Crytponix_: With enough RAM to fit all that data so it doesn't matter if it is SSD disk or not
2014-10-31 19:51:42
Crytponix_
salle: Ah OK. As we speak, I'm looking into XDR Rams for a card.
2014-10-31 19:52:14
Crytponix_
salle: The actual idea I have are two: 1) A high-speed card as a "Temporary Table" storage which can process very fast
2014-10-31 19:52:34
Crytponix_
salle: 2) A card with Flash and RAM and FPGA, which can act as one database
2014-10-31 19:53:12
salle
Crytponix_: Who is going to pay extra money for that when the RAM is already there? I won't
2014-10-31 19:53:59
Crytponix_
salle: The idea is that if you wait for CPU to take data from disk and put it in them ram, before it's sent to the card, then there is a timing penalty there. How important that is, I don't know
2014-10-31 19:54:17
Crytponix_
salle: Also sending data from RAM to Card will take time too. The overhead could be important
2014-10-31 19:54:35
Crytponix_
salle: or I'm far off?
2014-10-31 19:54:38
salle
Crytponix_: That's your problem :) You are investing lot of time to think about solving a problem which you don't know if even exists
2014-10-31 19:55:21
Crytponix_
salle: very true. I'm trying to get familiar with this first... Not easy. I saw the code, and one could easily find a lot of sources with more than 5,000 lines of code :D
2014-10-31 19:56:00
salle
Crytponix_: If you are talking about MariaDB code it is little more than 5000 lines
2014-10-31 19:56:24
Crytponix_
salle: Yes, what I mean was there are too many source files. And a lot of those source files are more than 5,000 lines
2014-10-31 19:56:40
Crytponix_
salle: *what I meant was...
2014-10-31 19:56:44
salle
:)
2014-10-31 20:52:41
otto
could mariadb.org webmasters please some warning to the view Galera is a fast synchronous multimaster wsrep provider (replication engine) for transactional databases and similar applications. For more information about wsrep API see http://launchpad.net/wsrep. For a description of Galera replication engine see http://galeracluster.com.</description>
2014-10-31 20:52:47
otto
</package>
2014-10-31 20:52:54
otto
ugh, wrong paste in buffer
2014-10-31 20:53:28
otto
could mariadb.org webmasters please some warning to the view https://downloads.mariadb.org/mariadb/repositories/#mirror=netinch&distro=Ubuntu&distro_release=trusty&version=10.1 what 10.1 is alpha and people should not add that repo to any production systems? thanks!
2014-11-01 00:12:30
Crytponix_
serg: Hello
2014-11-01 00:18:04
Crytponix_
serg: Hope I'm not being a pest (as columbo used to say :) )
2014-11-01 05:17:00
spstarr
Happy Mailman Day!
2014-11-01 05:28:42
synapt
Is there any particular way to flush a copied-in innodb table to a new mysql system where the table shows 'in use' from the old system?
2014-11-01 05:28:56
synapt
Trying to do anything with the table results in "doesn't exist in engine" error
2014-11-01 05:53:55
dcmwai
hello all, need some help on 2 node mariadb with galera cluster
2014-11-01 05:54:08
dcmwai
This is the error log on garbd http://dpaste.com/0Z0SZKA
2014-11-01 05:54:20
dcmwai
I configure my site as such http://www.sebastien-han.fr/blog/2012/04/01/mysql-multi-master-replication-with-galera/
2014-11-01 05:54:45
dcmwai
This is the guide I'm writting or in progress
2014-11-01 05:55:39
dcmwai
and I fail on doing this garbd -a gcomm://10.0.0.1:4567 -g my_wsrep_cluster -l /tmp/1.out -d
2014-11-01 05:55:46
dcmwai
Hum my guide https://wiki.gentoo.org/wiki/MariaDB_Galera_Cluster
2014-11-01 06:32:26
synapt
nobody around yet?
2014-11-01 12:41:33
DanGer
hello, I don't know if this is mariadb-specific or linux question, but I have 2 different linux systems, the old one running mysql and the new one running mariadb. When I run sysbench's oltp test on the old one, I see mysql is using a lot of CPU, while on the new one the CPU is almost idle. However I am getting 7x better results from the old server and I have no idea what is going on. What should I look at?