#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
2014-10-22 20:49:32
tyler_
I am getting a server version error when trying to run mysql_upgrade after removing mysql 5.0 from my server and installing mariadb 5.5 form my server. from what I have read it doesn't sound like that should be happening.
2014-10-22 21:01:02
kolbe
tyler_: what's the actual error you see?
2014-10-22 21:01:15
tyler_
Error: Server version (5.0.95) does not match with the version of
2014-10-22 21:01:15
tyler_
the server (5.5.40-MariaDB) with which this program was built/distributed. You can
2014-10-22 21:01:19
tyler_
use --skip-version-check to skip this check.
2014-10-22 21:02:06
kolbe
well if it's getting/giving that error you probably haven't really rmoved mysql 5.0 correctly
2014-10-22 21:02:22
kolbe
or maybe the mysql 5.0 server was still running when you removed it, and it was neveractually shut down?
2014-10-22 21:02:31
tyler_
hm that could be it one moment
2014-10-22 21:04:27
tyler_
it looks like the mysql process was still running after mysql was removed.
2014-10-22 21:04:34
kolbe
oops :)
2014-10-22 21:04:44
tyler_
thank you.
2014-10-22 21:05:07
tyler_
couldn't puzzle out why the server version was different then all the package information and mysql -V
2014-10-22 21:05:58
kolbe
:)
2014-10-23 10:58:21
Cryptonix
Hi All
2014-10-23 10:58:34
Cryptonix
Anyone there familiar with the internals of MariaDB code?
2014-10-23 10:59:14
Cryptonix
I need someone expertise for a Hardware project
2014-10-23 10:59:26
Cryptonix
A MariaDB Accelerator Card
2014-10-23 11:17:37
Cryptonix__
OK, here is the question: If were to make a card that accelerates SQL processing, what function would it help most to be integrated in the card? Sorting? Matching? Dictionarry Lookup? The card will be PCI-Express 8x or USB3, so the data transmission and reception time must be considered
2014-10-23 11:26:18
serg
Cryptonix__: that's too complex to answer on irc, better write an email. btw, did you hear of Kickfire?
2014-10-23 11:27:39
Cryptonix__
serg: Thanks for your comment. I'm trying to find a point of entry.
2014-10-23 11:27:57
Cryptonix__
serg: I have been told you're the best. Where can I send you an email to explain this?
2014-10-23 11:28:33
serg
Cryptonix__: I'm on #mysql too, I've seen that :) send an email to maria-developers@lists.launchpad.net
2014-10-23 11:29:14
serg
Cryptonix__: still, heard of Kickfire?
2014-10-23 11:29:49
Cryptonix__
serg: I just read about them.
2014-10-23 11:30:04
serg
okay. Just FYI
2014-10-23 11:31:00
Cryptonix__
serg: Well, I designed the Butterfly Labs cards. The same could be done here. I'm looking to make a small card that performs the acceleration. MySQL situation is a bit different, as it's memory-intensive and logic-intensive
2014-10-23 11:31:27
salle
Cryptonix__: "the same could be done here" is very brave statement :)
2014-10-23 11:31:49
serg
Cryptonix__: I don't really know what "Butterfly Labs cards" or "Butterfly Labs" but please mention that in the email, I'll google it up
2014-10-23 11:31:50
salle
Cryptonix__: Can you make card which accelerates everything anyone would ever want to do with a computer?
2014-10-23 11:31:53
Cryptonix__
salle: Well, we did that for SHA256 (20,000 fold speed increase compared to intel XEON)
2014-10-23 11:32:11
archivist_herron
some stuff is easy
2014-10-23 11:32:40
Cryptonix__
salle: Well you can't accelerate everything. You dedicate a large number of transistors to a specific task
2014-10-23 11:32:55
Cryptonix__
salle: that would accelerate that particular task
2014-10-23 11:33:13
salle
Cryptonix__: That's exactly the problem with SQL. It is too generic with wide range of tasks it is used for
2014-10-23 11:33:50
salle
Cryptonix__: Who cares for example about sorting accelerator if he never uses any queries which sort more than 10 rows?
2014-10-23 11:33:54
Cryptonix__
salle: Yes. But I'm sure if some statistics are run, it'll be more visible that most of the time of the CPU is being consumed by sorting, or adding, or ... ? That's what can be integrated in a separate card
2014-10-23 11:34:10
serg
salle: it wouldn't be the first attempt to use hardware acceleration MySQL/MariaDB. May be eventually somebody comes with an idea that succeeds
2014-10-23 11:34:23
salle
Cryptonix__: You will fail to find such generic statistics
2014-10-23 11:34:26
Cryptonix__
salle: Large organizations do need that. I've been in a bank IT and Stock-Exchange as well. 200,000 rows is something to begin with there
2014-10-23 11:34:34
salle
serg: Maybe!
2014-10-23 11:34:59
salle
Cryptonix__: Sorting 200 000 rows frequently would be a design mistake :)
2014-10-23 11:35:18
Cryptonix__
salle: Yes, and design-mistakes are what banks are all about. You wouldn't believe what you would find there
2014-10-23 11:36:20
Cryptonix__
salle: They are way behind in correct-design. Banks are run by contractors who on average stay there like 6months to a year. After one year, no one knows what the previous designer has done. This is where things start to look funny
2014-10-23 11:36:51
salle
Cryptonix__: Let me make it clear: You *can* design such hardware which helps accelerating MySQL or any other RDBMS for very concrete application
2014-10-23 11:37:26
Cryptonix__
salle: Yes. However, once certain features are available on hardware very fast, queries would be re-written to take advantage of those features.
2014-10-23 11:37:52
salle
Cryptonix__: You can go at some company and work with them to evaluate their needs are and design such hardware which accelerates their specific application running against their specific data. That's absolutely possible
2014-10-23 11:38:00
Cryptonix__
salle: This adventure brings certain very high-speed features, and evolves around designers changing their queries to take advantage of those. Pretty much like parallel computing
2014-10-23 11:38:12
salle
Cryptonix__: You can not design something which could help everyone
2014-10-23 11:38:44
Cryptonix__
salle: I'm also aware of that as well. Right now, what I'm looking for is bottle-necks on CPUs for query processing
2014-10-23 11:39:06
Cryptonix__
salle: What the bottle-necks are will determine the fate of the project
2014-10-23 11:39:23
salle
Cryptonix__: What you can look at are some badly designed, but often used frameworks which use MySQL and accelerate them.
2014-10-23 11:40:02
salle
serg: Bottlenecks on CPU for query processing? What this could possibly mean with MySQL?
2014-10-23 11:40:43
jkavalik
evaluating lot of different and maybe exotic plans for more complex queries?
2014-10-23 11:41:45
salle
Cryptonix__: Sorting huge tables is actually disk bound not cpu bound
2014-10-23 11:42:07
Cryptonix__
salle: Here is an example: A query requires two columns to be added and returned. There are like 800 rows to be returned, this means 800 additions (which are unrelated to each other). A parallel design could do the 800 additions in like 8 cycles. A CPU has to do it in 800 cycles, unless some SSE feature is used
2014-10-23 11:43:55
salle
Cryptonix__: That will take miliseconds on modern hardware and the rest of the query will take more
2014-10-23 11:43:57
Cryptonix__
salle: this reduces the processing time from 800 x 0.5ns (for a 2GH CPU) = 400ns to 8x2ns (500MHz ASIC or FPGA) which equals 16ns
2014-10-23 11:44:20
salle
Cryptonix__: Please try few complex queries with SHOW PROFILE for them
2014-10-23 11:44:47
Cryptonix__
salle: So in your opinion the bottle-neck is not the CPU but data fetching from Harddisk?
2014-10-23 11:44:54
Naktibalda
but it will be 1% improvement for a total time of execution
2014-10-23 11:45:58
Cryptonix__
Naktibalda: Yes in the given example it's like that. I'm trying to share my vision and have your vision to see where they cross
2014-10-23 11:46:01
salle
Naktibalda: I'd say 0.1% or less
2014-10-23 11:46:40
salle
Cryptonix__: With MySQL CPU is rarely the bottleneck and when it is it is either due to misconfiguration or bad design or it can not be accelerated
2014-10-23 11:47:02
salle
Cryptonix__: You can not accelerate with hardware high locking contention for example
2014-10-23 11:47:47
salle
Cryptonix__: I am trying to explain you whatever acceleration you aim for it will be useful in very narrow niche
2014-10-23 11:49:19
Cryptonix__
salle: I'm searching for that narrow niche actually. The environment this will be run at is like minimum 200,000 rows or even 1 million row
2014-10-23 11:49:25
salle
Cryptonix__: You can for example do it for spatial features and there probably you can accelerate a lot, but spatial features are used by less than 1% of users
2014-10-23 11:49:57
salle
Cryptonix__: Or maybe you can accelerate fulltext indexes somehow
2014-10-23 11:50:13
salle
Cryptonix__: That would be useful for lot more users
2014-10-23 11:50:40
Cryptonix__
salle: hmm, Spatial you mean geometry (Sine, cosine, etc)? Also let me look-up fulltext indexs
2014-10-23 11:50:51
salle
Cryptonix__: However if you want to come with something which will be useful for everyone forget about it
2014-10-23 11:51:10
salle
Cryptonix__: Nope. I mean GIS related features
2014-10-23 11:51:51
Cryptonix__
salle: Yes something to help everyone cannot be found. It will start as specialized, and it may grow from there as community demands features
2014-10-23 11:52:26
Cryptonix__
salle: just wondering, could you give me an idea of fulltext indexes?
2014-10-23 11:52:35
salle
heh
2014-10-23 11:52:52
Cryptonix__
salle: Also string-processing is another thing that could be done
2014-10-23 11:54:01
Cryptonix__
salle: OK the fulltext indexes is something that can be done definitely. FPGA/ASIC can be used to scale it up
2014-10-23 11:54:47
salle
Cryptonix__: I'm not the one to impress with that :) I believe full text indexing in RDBMs is mistake :)
2014-10-23 11:55:11
salle
serg: You wrote most of Fulltext in MyISAM didn't you?
2014-10-23 11:55:45
Cryptonix__
salle: But in general, string search and manipulation is something that can be paralleled. It does eat up CPU time in large queries
2014-10-23 11:56:30
Naktibalda
in general it is recommended to avoid doing that in SQL
2014-10-23 11:57:25
Cryptonix__
Naktibalda: Yes very true. But businesses use that immensely. It comes from point of contact of staff and ERPs. EDI Systems use it a lot (electronic data interchange between corporations)
2014-10-23 11:58:38
Naktibalda
how many banks use mysql?
2014-10-23 11:58:48
salle
Cryptonix__: Back to my point: You can do it for *specific* bussiness case
2014-10-23 11:58:49
lefred
lefred dreams of world where all things recommended to be avoided in MySQL won't be massively used :-D
2014-10-23 11:58:52
Cryptonix__
Naktibalda: A lot.
2014-10-23 11:59:23
salle
lefred: Shht .. lot of us will lose our jobs in such world :)
2014-10-23 11:59:30
lefred
salle: ;-)
2014-10-23 11:59:33
Cryptonix__
Naktibalda: As new contractors or teams come in a bank or insurance business, and where they have the right to use their own tools, you see diversity
2014-10-23 12:00:35
Cryptonix__
Naktibalda: You will find MS-SQL, Sybase, Postgre-SQL, MySQL and a bunch of other databases in the SAME company. Making this card will give MariaDB or MySQL a cutting edge and an incentive for everyone to switch
2014-10-23 12:00:46
Cryptonix__
Naktibalda: It's a win-win situation
2014-10-23 12:00:54
salle
Cryptonix__: I don't quite believe you lot of banks use MySQL and I also do have reasons to believe I know that much better than you
2014-10-23 12:01:09
salle
Cryptonix__: Lot of companies yes, but not banks
2014-10-23 12:01:31
Cryptonix__
salle: I've been in 2 comapanies in France, a very huge bank, an insurance corporation, and a logistic corporation
2014-10-23 12:01:54
Cryptonix__
Logistic one sticked with MS-SQL. They were small however
2014-10-23 12:01:56
salle
Cryptonix__: If making such card was possible there would be many competitors on that market already. SQL is around for more than 30 years in production systems
2014-10-23 12:02:19
Cryptonix__
the two other companies used MySQL and Postgre SQL. It was a 200 million euro french company, the other was smaller.
2014-10-23 12:02:42
salle
Cryptonix__: One small bank != many banks
2014-10-23 12:03:04
Cryptonix__
salle: The second largest bank in France
2014-10-23 12:03:33
salle
Cryptonix__: The second largest bank in France uses MySQL for critical bussiness application?
2014-10-23 12:03:38
Cryptonix__
salle: I also know people worked in europes largest bank (which I cannot name). They were very into Java.
2014-10-23 12:04:16
Cryptonix__
salle: There is no single critical database there. There are many. Too many teams there, hence you see too many different approaches.
2014-10-23 12:04:30
salle
Cryptonix__: Lets not talk about each of us knows please ...
2014-10-23 12:05:09
Cryptonix__
salle: Just wanted to give you an idea of the diversity and inefficiency that won't look apparent at all
2014-10-23 12:05:42
salle
Cryptonix__: Please don't teach me about who and how uses MySQL
2014-10-23 12:06:13
pomyk|w
Cryptonix__: don't bother with sql
2014-10-23 12:07:39
Cryptonix__
salle: I never intended that honestly, sorry if I said something that looked that way
2014-10-23 12:08:47
Cryptonix__
salle: Thanks for your help guys, I much appreciate it
2014-10-23 12:09:08
Cryptonix__
Naktibalda: Thanks for the info.
2014-10-23 12:09:19
Cryptonix__
serg: Thanks and I wil be sending an email, explaining this in full
2014-10-23 12:14:19
salle
Cryptonix__: Let me give you an example. Here is what SHOW PROFILE retunrns for rather slow query: 259 rows in set (27.75 sec)
2014-10-23 12:14:38
salle
Cryptonix__: http://pastie.org/9669954 Which part of this you are going to accelerate and how much it is going to matter?
2014-10-23 12:15:11
salle
Cryptonix__: Note what % of the time is spent for sorting the result
2014-10-23 12:16:30
erkules
erkules hugs salle
2014-10-23 12:17:14
salle
Cryptonix__: Even if you accelerate the sorting by factor of 10000 would it matter enough to clarify purchasing separate hardware for that?
2014-10-23 12:18:08
Naktibalda
isn't that "sending data" a misleading name?
2014-10-23 12:18:22
salle
Naktibalda: yes it is misleading, but it can't be accelerated
2014-10-23 12:18:38
salle
Naktibalda: Fetching rows is not CPU bound
2014-10-23 12:18:45
Naktibalda
why is it sorting before execution?
2014-10-23 12:19:34
salle
Naktibalda: perhaps because it is type=index query
2014-10-23 12:21:10
salle
Naktibalda: I don't think they are in precise order by the way
2014-10-23 12:25:31
salle
Naktibalda: Here is better one this time with filesort: http://pastie.org/9669975
2014-10-23 12:26:01
salle
Naktibalda: Creating sort index | 0.000551 is now after Sending data | 27.771263
2014-10-23 12:27:14
Naktibalda
there is an obvious candidate for optimization :)
2014-10-23 12:27:34
Naktibalda
or additional profiling
2014-10-23 12:27:43
Naktibalda
other parts are irrelevant
2014-10-23 12:28:33
salle
Naktibalda: My bad .. that was from MySQL. MariaDB 10 is more reasonable with SHOW PROFILE OUTPUT; same query, smaller table http://pastie.org/9669982
2014-10-23 12:29:07
salle
Naktibalda: My point was to show Cryptonix__ random example which shows there is very little to be achieved by hardware acceleration
2014-10-23 12:29:54
salle
Naktibalda: No room for optimization by the way because the query is as simple as: select x, count(*) from t group by x order by count(*);
2014-10-23 12:30:52
Cryptonix__
salle: Thanks for the example. My intention is to build a card with specific features, and evolve the code into using those features. It does seem nearly impossible to make an accelerator that works for all in any situation...
2014-10-23 12:31:03
salle
Naktibalda: Unless indeed some denormalization can be considered :)
2014-10-23 12:31:37
Naktibalda
but it would eliminate calculation from query
2014-10-23 12:33:44
salle
Cryptonix__: I am afraid it is even worst than that. You can accelerate only very specific cases. It could be for example possible to create accelerator for certain widely used software wich uses database with bad design. There are plenty of them
2014-10-23 12:34:28
salle
Cryptonix__: Look at the last one: http://pastie.org/9669982
2014-10-23 12:35:12
salle
Cryptonix__: Here the most time is spent for copying data into the memory used for sorting (memory based temp table) not for the sorting itself
2014-10-23 12:36:07
Cryptonix__
salle: ah thanks a very good example, thanks
2014-10-23 12:36:15
Cryptonix__
going to groceries, be back in a jiffy
2014-10-23 12:36:21
Cryptonix__
:)
2014-10-23 12:36:57
salle
Cryptonix__: You can accelerate the sorting part, but it is 3000 times faster anyway than the time used to copy required data and this part you can't accelerate with hardware add-on card
2014-10-23 12:39:12
salle
Cryptonix__: And even if I add some complex math expressions in SELECT part it will not change much.
2014-10-23 12:41:21
Cryptonix__
hmm, another idea i had in mind was a 2TB/s SSD designed to connect to PCI Express 8x.
2014-10-23 12:41:43
Cryptonix__
But I think with existance of RAM disks, not many people would be interested. Also SSD does wear out
2014-10-23 12:44:16
salle
Cryptonix__: By the way neighter MariaDB nor MySQL can parralelize the execution of single query so far
2014-10-23 12:45:14
Cryptonix__
then the only way to accelerate is to have faster storage I assume?
2014-10-23 12:45:22
Cryptonix__
faster access to data, rather than faster processing
2014-10-23 12:46:04
Cryptonix__
"Copying to tmp table" is copy data from Disk to RAM or RAM to RAM?
2014-10-23 12:49:11
lefred
Cryptonix__: it could be ram to disk too
2014-10-23 12:50:51
Cryptonix__
salle: I also see a 27 seconds for "Sending Data" in one of those pasties.
2014-10-23 12:51:07
salle
Cryptonix__: If query is properly optimized "Copying to tmp table" is RAM to RAM
2014-10-23 12:51:11
Cryptonix__
lefred: yes it could be. 1 second is a long time though
2014-10-23 12:51:33
salle
Cryptonix__: Sendind data is misleading because it means lot more
2014-10-23 12:51:51
Cryptonix__
salle: I see, well I'm learning along
2014-10-23 12:52:56
salle
Cryptonix__: For example Sending data includes processing everything in SELECT part along with fetching the data from .. wherever it is fetched which varies depending on query
2014-10-23 12:53:22
salle
Cryptonix__: It could be index structure alone or index + data or data alone
2014-10-23 12:53:36
salle
Cryptonix__: .. which means it depends on storage engine too
2014-10-23 12:54:58
salle
Cryptonix__: If you dig deep into the InnoDB storage engine which is the most widely used nowadays perhaps you can see something which can be accelerated by hardware there and that can be a market hit :)
2014-10-23 12:55:40
salle
Cryptonix__: Like optimizing the notoriously slow SELECT COUNT(*) FROM tbl; without WHERE clause
2014-10-23 12:56:05
Naktibalda
but it isn't a hardware issue!
2014-10-23 12:56:30
salle
Naktibalda: Absolutely :) Nor it is CPU bound
2014-10-23 12:58:20
salle
Naktibalda: Maybe he can find a way to keep all index statistics up to date all the time doing it in the background instead of being approximate as they are now. That will help every single query ... a little. I'd say 0.01% in general :)
2014-10-23 13:00:47
salle
Naktibalda: Or rewrite stored routines from scratch and design hardware accelerator for them ..
2014-10-23 13:08:25
Cryptonix__
salle: I'll write the "SELECT COUNT(*)" down in my notes
2014-10-23 13:08:53
Cryptonix__
salle, naktibalda: By the looks of it, this SQL is mostly Memory work rather than logic work. Correct me if I'm wrong
2014-10-23 13:09:30
Cryptonix__
I also believe a good portion of InnoDB Engine has to be re-written to take advantage of any hardware acceleration.
2014-10-23 13:09:30
salle
Cryptonix__: Depends on query
2014-10-23 13:09:50
Naktibalda
myisam storage engine has the count stored in metadata, innodb actually has to process all rows, because the result can be different for different transactions
2014-10-23 13:09:52
salle
Cryptonix__: Rewriting any portion of code is dead trap for you :)
2014-10-23 13:10:36
salle
Cryptonix__: You have to constantly keep it up to date so you have to maintain your own fork or something similar
2014-10-23 13:11:21
Cryptonix__
salle: Well, in scenario of playing or filtering 200,000 records, it seems memory-fertching would take more time rather than processing. Of course, data has to reach CPU cache in smaller portions. Hence DISK -> RAM -> CPU -> RAM -> CPU -> RAM -> Result
2014-10-23 13:11:50
Cryptonix__
salle: It would be a bad trap, considering the fact that I have never done it before either :)
2014-10-23 13:12:06
Cryptonix__
Hence community supporting the project will ensure it's survivability
2014-10-23 13:13:01
salle
Cryptonix__: Remember I pointed to the fact that SQL is around for more than 30 years in production and asked how many hardware accelerators exist for it
2014-10-23 13:14:07
Cryptonix__
salle: It's a very valid point. There must be a reason. However, some hardware has been built for it (directly or indirectly). Ram Disks would be one I think. I think they had the SQL in the back their mind when designing the Ram Disk
2014-10-23 13:14:10
salle
Cryptonix__: As a language it is 40 years old, but it wasn't widely used before 1980s in production
2014-10-23 13:14:54
salle
Cryptonix__: Did you read aboit Kikstarter?
2014-10-23 13:15:10
salle
sorry
2014-10-23 13:15:13
salle
Kickfire
2014-10-23 13:17:18
Cryptonix__
salle: Yes I did. I see it as an attempt with venture capital funding. By the looks of it, however, they weren't very successful (I get the feeling as I have never heard from them in companies I worked in)
2014-10-23 13:17:23
Naktibalda
Naktibalda saw this in the search results before: http://ocz.com/enterprise/zd-xl-sql-accelerator-1.5-pcie-ssd
2014-10-23 13:17:32
jkavalik
bit offtopic - but there are special processors running lisp, java bytecode or some similar stuff, anyone want to hardcode sql processing into special chip with enough cache, lots of ram and direct connection to big ssd ? :)
2014-10-23 13:19:57
salle
jkavalik: Which SQL dialect? :)
2014-10-23 13:20:12
Cryptonix__
Naktibalda: That's a good example. I need to dig into it
2014-10-23 13:20:40
jkavalik
salle, pick one ;) something like mysql/maria server "on a chip"
2014-10-23 13:21:09
salle
jkavalik: How do you fix bugs with such chip?
2014-10-23 13:21:57
Cryptonix__
jkavalik: a full SQL server isn't worth implementation on a chip. It's too complicated, requires too many transistors, no upgrades possible
2014-10-23 13:22:17
jkavalik
salle, how do they fix bug in those java/lisp? but suppose they are A LOT simpler
2014-10-23 13:22:32
jkavalik
Cryptonix__, I see
2014-10-23 13:22:36
Cryptonix__
jkavalik: Also the problem with debugging as well. A co-processor or an accelerator would fit better
2014-10-23 15:06:06
fossxplorer
Any one worked on Ansible playbook for Centos 7 and Maria 10?
2014-10-23 15:06:14
fossxplorer
or should i work on one?
2014-10-23 15:17:44
erkules
what is the purpose of the playbook?
2014-10-23 15:18:31
erkules
erkules knows ansible just wonders what fossxplorer would like to achieve
2014-10-23 15:19:52
fossxplorer
i'm modifying the LAMP_simple from https://github.com/ansible/ansible-examples to include MariaDB with some updated PHP stuff.
2014-10-23 15:20:40
fossxplorer
the db playbook should install MariaDB instead of MySQL to start with, but i'll keep improving it a better config file template etc
2014-10-23 15:21:36
fossxplorer
erkules, IIRC, you mentioned you gave talk about Ansible + Maria a while back
2014-10-23 15:22:47
tanj
fossxplorer: I have one somewhere
2014-10-23 15:23:00
tanj
for CentOS 6, but should work with 7 as well...
2014-10-23 15:23:06
erkules
fossxplorer: even with galera
2014-10-23 15:23:55
erkules
But that kind of playbooks is quite boring :)
2014-10-23 15:24:05
erkules
But that kind of playbooks are quite boring :)
2014-10-23 15:24:26
tyler_
should be pretty simple. my playbook just installs maria based on the os version
2014-10-23 15:25:17
tyler_
if you are using roles just switch out the database role from mysql to maria and you are good
2014-10-23 15:25:30
tanj
yeah much boring indeed :)
2014-10-23 15:25:56
tyler_
i like boring. it means things are working right.
2014-10-23 15:26:18
erkules
right
2014-10-23 15:26:42
fossxplorer
erkules, ah, yeah Galera did you mention
2014-10-23 15:27:35
fossxplorer
hehe, boring for someone, but i'm new to Ansible, kinda excited to get it working
2014-10-23 15:27:56
fossxplorer
Also, i like to use Marias repo and not the ones from OS
2014-10-23 15:28:20
erkules
yeah if it is your first configmanagement you choosed the right one
2014-10-23 15:28:47
fossxplorer
I tried Puppet for some months..good but quite complex
2014-10-23 15:29:12
erkules
just to rephrase tyler_ : The architecture of ansible is boring
2014-10-23 15:29:17
erkules
and boring is good :)
2014-10-23 15:29:47
fossxplorer
hmm, you guys like Ansible?
2014-10-23 15:29:56
tyler_
I do like it. pretty handy.
2014-10-23 15:29:59
fossxplorer
and prefer it over puppet?
2014-10-23 15:30:02
erkules
In encourage to use it
2014-10-23 15:30:09
erkules
no
2014-10-23 15:30:15
tyler_
i am using it to upgrade a bunch of mysql 5.0 boxes to 5.5
2014-10-23 15:30:19
tyler_
*maria 5.5
2014-10-23 15:30:20
fossxplorer
Cool. I just hope not to have chosen a wrong CM tool
2014-10-23 15:30:33
fossxplorer
Cool
2014-10-23 15:31:38
fossxplorer
I need to setup servers with Nginx, MariaDB, PHP 5.5. Hope Ansible will be able to fix it for me soon
2014-10-23 15:32:38
erkules
you will prevail :)
2014-10-23 15:33:12
tanj
Ansible can't be wrong :)
2014-10-23 15:34:16
fossxplorer
Hehe, thx, give me motivation :)
2014-10-23 15:34:17
erkules
rofl. Which indeed is wrong. thats the good thing about puppet. It makes sure every resource is defined one
2014-10-23 15:34:22
erkules
once
2014-10-23 15:34:38
erkules
erkules loves to take the fun out of ...
2014-10-23 15:35:13
fossxplorer
Hehe. I understand i need to put some time & effort into creating these playbooks, then to benefit from them
2014-10-23 15:45:23
pablo_
Hello,.. I'm changing mariadb code.... Is there some function that I can run to find a thread and attach to it in with a debugger mariadb? : )
2014-10-23 16:50:57
xenol
tanj: do you know ansible-shell? :)
2014-10-23 17:12:32
tanj
xenol: nope
2014-10-23 17:12:35
tanj
pablo_: use gdb
2014-10-23 19:21:58
oozbooz
we project our MariaDB will reach size of 1TB (support 10 clients )by the end of the year, is there a guide on how to spec server or VM/Container?
2014-10-23 19:24:29
thumbs
you're considering a VM?
2014-10-23 19:29:26
kolbe
dbart: is it possible for you to kick off some custom build of our 64-bit generic linux binary with performance schema disabled?
2014-10-23 19:40:53
oozbooz
thumbs, a container..
2014-10-23 19:42:03
oozbooz
when spec says DB size is 1TB... does it translates into 1TB+ RAM?
2014-10-23 19:42:34
thumbs
oozbooz: you don't have to fit the whole thing in the buffer pool
2014-10-23 19:43:24
oozbooz
ok... how one estimates RAM requirements for 1TB DB
2014-10-23 20:05:40
kolbe
elenst: ping
2014-10-23 20:05:54
elenst
kolbe: pong
2014-10-23 20:06:07
kolbe
elenst: do you know anything about buildbot or the MariaDB build process?
2014-10-23 20:06:17
elenst
kolbe: some
2014-10-23 20:06:25
elenst
kolbe: what do you want to know?
2014-10-23 20:06:40
kolbe
i want to get one of our Generic Linux x86-64 (not glibc 2.14) binaries built but with performance schema completely disabled and left out of it
2014-10-23 20:06:51
elenst
ok..
2014-10-23 20:06:59
kolbe
do you know how i could go about that?
2014-10-23 20:07:06
kolbe
or who would be the best person to talk to?
2014-10-23 20:07:18
elenst
and how is it connected to buildbot? you want it to be done in buildbot?
2014-10-23 20:07:32
kolbe
i don't care how it's done
2014-10-23 20:07:38
kolbe
perhaps it has no connection at all to buildbot
2014-10-23 20:08:16
elenst
ok
2014-10-23 20:08:30
elenst
kolbe: that's the exact steps buildbot makes to create the bintar: http://buildbot.askmonty.org/buildbot/builders/kvm-bintar-centos5-amd64/builds/798/steps/compile/logs/stdio
2014-10-23 20:08:34
elenst
(for 5.5)
2014-10-23 20:08:58
kolbe
well, ok
2014-10-23 20:09:12
kolbe
i'm not really sure what i can do with that information to get a build produced on that same machine
2014-10-23 20:09:12
elenst
the simplest way is to perform the same steps but add an option to exlude perfschema
2014-10-23 20:09:20
kolbe
(and i need it for 10.0.14 anyway)
2014-10-23 20:09:55
elenst
kolbe: okay, lets put it another way :) do you need to know how it's done, or do you need it done?
2014-10-23 20:10:14
kolbe
i don't care at all how it's done, i just need it to happen :)
2014-10-23 20:10:41
elenst
kolbe: 10.0.14 release sources?
2014-10-23 20:10:46
elenst
no custom patches?
2014-10-23 20:11:22
kolbe
no, nothing custom except whatever cmake option completely disables and excludes performance schema