RSS: linuxkernel: 2.6.27-rc4-git4: stable snapshot <http://www.kernel.org/kdist/fragments/stable_snapshot.html>

gn8 all

*sniff* R.I.P. query.php

*sing* Alles hat ein Ende, nur die Wurst hat 2... ;)

?

hi, are there any toolserver admins around?

_mary_kate_ > you around?

cron is currently broken on nightshade

'date' returns the correct date, but when it's run under cron, it's two hours fast... so all jobs run two hours early :|

toolserver-l: 03Kalan * 10/: Re: [Toolserver-l] PHP error handling changes

RSS: sunsecurity: Sun Alert 241166 SUN ALERT WEEKLY SUMMARY REPORT - Week of 17-Aug-2008 to 23-Aug-2008 <http://blogs.sun.com/security/entry/sun_alert_241166_sun_alert> || Sun Alert 241066 A Security Vulnerability in the Solaris NFS Kernel Module May Lead to a System Panic, Resulting in a Denial of Service (DoS) <http://blogs.sun.com/security/entry/sun_alert_241066_a_security>

@replag

WaRpAtH: s1: 1s [no change]; s2: 1s [-0.00 sec/sec]; s3: 56m 18s [+0.04 sec/sec] (21h 27m 40s sample)

@replag

_mary_kate_: s1: 2s [+0.00 sec/sec]; s2: 33s [+0.04 sec/sec]; s3: 1h 7m 7s [+0.74 sec/sec] (14m 34s sample)

@replag

Werdna: s1: 23s [+0.06 sec/sec]; s2: 3s [-0.08 sec/sec]; s3: 1h 12m 36s [+0.86 sec/sec] (6m 21s sample)

:(

_mary_kate_: could you do a favour?

depends

_mary_kate_: do you have an x86_64 computer running linux handy?

no, but you do

I can't compile AFP on my server, it's too memory-intensive and I haven't gotten to asking for more ram yet

my only amd64 machine is my desktop, which runs solaris

Werdna: so use nightshade? :)

toolserver maybe?

werdna@nightshade:~$ uname -a

Linux nightshade 2.6.26.3-hemlock36-fp #1 SMP Fri Aug 22 03:36:21 UTC 2008 x86_64 GNU/Linux

sweet!

thanks.

@replag

Werdna: s1: 0s [-0.01 sec/sec]; s2: 1s [-0.00 sec/sec]; s3: 45m 45s [-0.62 sec/sec] (43m 23s sample)

@replag

Werdna: s1: 2s [+0.00 sec/sec]; s2: 0s [-0.00 sec/sec]; s3: 54m 11s [+0.17 sec/sec] (48m 50s sample)

3(created) [ST-46] How can I get the Dotmasters page back up ?; ST47's tools; Bug <10https://jira.toolserver.org/browse/ST-46> (Sandra)

_mary_kate_: what should I do if I get 500 error on my PHP tool?

switchboard error O_O

VasilievVV: does your tool take more than 10 minutes to generate output?

_mary_kate_: from 15 seconds to 3 minutes, I think

VasilievVV: do you have a URL that produces the error?

_mary_kate_: last time I checked, it didn't produce any errors

But thanks anyway for attention

hm

VasilievVV: if you happen to be around when someone reports an error, look in error_log for whether it generates a fastcgi error (like this: (4)Interrupted system call: FastCGI: comm with server "/var/www/cgi/php" aborted: select() failed)

it'll have the client's IP in

there are quite a few of those and while i think most of them are from broken scripts, the error is so vague it's hard to be sure

(so far, no one has reported a broken script that definitely wasn't caused by a timeout...)

error_log?

/var/log/apache2/error.log

[Tue Aug 26 12:31:48 2008] [error] [client 213.181.10.212] (4)Interrupted system call: FastCGI: comm with server "/var/www/cgi/php" aborted: select() failed

[Tue Aug 26 12:31:48 2008] [error] [client 213.181.10.212] FastCGI: incomplete headers (0 bytes) received from server "/var/www/cgi/php"

toolserver-l: 03Carl Beckhorn * 10/: Re: [Toolserver-l] PHP error handling changes

@replag

erwin85: s1: 1s [-0.00 sec/sec]; s2: 1s [+0.00 sec/sec]; s3: 1h 3m 0s [+0.07 sec/sec] (2h 2m 19s sample)

LOL, I've never seen a 500 error tagged onto the end of a page :P

mod_fastcgi does that

http://i38.tinypic.com/1qogox.png

ahh

SQLDb, mine gets that occasionally

http://toolserver.org/~soxred93/userstats/ <- This one doesn't

SQLDb goes to crash TheLetterX's tool

_mary_kate_: mounting hemlock's /var/log via nfs on nightshade would be convenient...

let's see how it likes processing SmackBot :P

Duesentrieb: i've never found a need for it... just edit scripts on hemlock if you want to edit, save, ^Z, tail, fg, edit...

_mary_kate_: well, we tell people to log into nightshade, not hemlock... anyway, there's demand for it, see the last two messages on the ML

Yeah, I would kinda like it too

it would just be convenient

I just plain try not to login to hemlock

by the way... how big are the pipes between login, stable and www?

GigE, If memory serves from what I've heard here :P

yea, i seem to recall that too

yes

_mary_kate_: so, would you set up an nfs mount? i have never done that :)

i'm trying to think of a reason not to, i really hate the idea ;)

3(commented) [DBQ-36] enwiki list of chatter <10https://jira.toolserver.org/browse/DBQ-36> (zeyi He)

apparently, nightshade's clock went back 2 hours somewhere yesterday

I received both of this morning's cron mails 2 hours earlier than usual :)

i think it's just cron

hmm

restarted it, should be fixed now

okay ^^

hello

hello fale

hi DaBPunkt

toolserver-l: 03River Tarnell * 10/: [Toolserver-l] [stable] proxy server maintenance,

Has there been any changes to the toolserver recently? I'm getting a lot of 500-internal-server-error messages on various longer-running scripts.

Pathoschild: see ML

Pathoschild: how long?

I've only noticed it over the last two or so days.

i mean how long do they run

DaBPunkt: do you know if boost is already installed?

_mary_kate_: They seem to hang around the 10-30 second point, then output an error.

fale: no idea - but there are severall debian-packages: so installation should be no problem

DaBPunkt: I think is not as you think, that packages are very old

Pathoschild: example?

DaBPunkt: hm, where's this problem discussed on the ML?

_mary_kate_: <http://toolserver.org/.../sul_log.php?limit=500&show_locked=1&show_hidden=1>, <http://toolserver.org/~luxo/contributions/contributions.php?user=Kate>.

oh, you mean someone else's tools

It also occurs on one of mine, although only in its most inefficient mode: <http://toolserver.org/.../?prefix=&category=Aragon%C3%A9s&database=sourceswiki-p>

i'm going to replace hemlock's mod_fastcgi with a slightly modified one

okay, done. either it'll help the problem a bit or it'll break horribly ;)

Woo, that is lightening fast.

The problem is gone; thanks. Will it stay this fast, or will it bog down when tools come back online? ;)

the speed is unlikely to change

it's probably just fast because apache was stopped for a few minutes

:D

_mary_kate_: so you found something?

i modified it so if select() returns EINTR (Interrupted system call) it will try again, instead of giving up

that's how select is usually used, i don't know why it didn't do that already

but the code is kind of hairy, it's hard to understand

there seem to be fewer fcgi errors in the log now

yea, let's wait and see if it improves for people.

if that did it, you should probably push that patch to the maintainers of fastcgi. or at least ask them why they don't do it that way :)

yes

now my timeout test script produces a switchboard error (as expected) instead of an internal server error. so something is probably happier

balasyum: Are you here?

here I am, MinuteElectron

Have you got a toolserver account yet?

i'm going to restart switchboard to reset the error count

No, MinuteElectron and I don't think I will ever.

ok

Flyingparchment and/or the other admins stick the account to revealing real name.

Duesentrieb: switchboard is recording fewer request errors now. 1 failed request out of 514. (used to be more like 1/100)

nice

balasyum: Wikimedia Germany makes the policies for granting accounts.

I did not blame anyone.

toolserver-l: 03River Tarnell * 10/: [Toolserver-l] internal server errors

MinuteElectron, if all goes well, I may soon have my cluenet account, so I can do tasks that don't need access to the wikimedia db.

My tool is a mystery: if I *don't* add a user name i become a 500 error, if i add a user name (so it should load longer) I become no error :?

?

So many 500 errors recently

can you login in on the toolserver? :S

I cannot upload files :S

maybe is only a my problem but..

where are you trying to log in?

_mary_kate_: login.toolserver.org

here it works.

what is your IP?

_mary_kate_: sec

_mary_kate_: 79.7.247.8

uhm, that I shoulded post in query.. -.-

well, np

*attack* *hack* :P

Luxo: :)

you failed to log in so many times your IP was blacklisted. try again now

ooops

_mary_kate_: sorry :)

Filnik retries

_mary_kate_: seems a my problem so :S

uuuhm

_mary_kate_: If I try with the console:

fil@deeder:~$ ssh toolserver

ssh_exchange_identification: Connection closed by remote host

*shell

DaBPunkt: can you make denyhosts stop being retarded? every time i remove an IP, it puts it back immediately

_mary_kate_: ehm.. :)

wait for dab

_mary_kate_: ahn, ok sorry :)

_mary_kate_: I know. Workarround: Add the IP temporaly to deny.allow

host.allow

okay.. feel free to that for filnik ;)

sek

_mary_kate_: ehrm, can I try now? :)

Filnik feels noob :(

Filnik: please try

Filnik: why is it you seem to have an 'svn co' running all the time? you know you can update an existing copy without checking out a new one, right?

_mary_kate_: uhuh?

_mary_kate_: it updates every 2-3 minutes :S

_mary_kate_: and how should I do to do it automatically?

(also a link is welcome :) )

just run 'svn up' from inside the existing source instead of running another svn co

(logged in with the shell, let's try with the other)

_mary_kate_: should I do it only once or in the crontab? (thanks for the help btw! :) )

do it whenever you want to update your copy of whatever it is you're updating

_mary_kate_: and what's the difference with the other way I do it? :)

svn co retrieves the entire source every time

ahn

:S

_mary_kate_: fish://login.toolserver.org/home/filnik/ <-- where's the error? :(

no idea what is fish://

ahn, ehm

Aug 26 17:09:07 nightshade sshd[25243]: Invalid user fil from 79.7.247.8

ahn

-_-

sorry

_mary_kate_: works now! :)

for svn up

_mary_kate_:

filnik@nightshade:~/pywikipedia$ svn up

svn: Working copy '.' locked

svn: run 'svn cleanup' to remove locks (type 'svn help cleanup' for details)

and svn cleanup seems taking too much time to resolve, seems going in standbye or so

probably a 'co' is already running, or was interrupted

ahn

_mary_kate_: it goes in standbye and nothing else :(

(svn cleanup, I mean)

could someone help me with a query please?

Just ask.

i want a list of articles in a given category on enwiki that don't have an interwiki link to arwiki

is it possible?

Yes, I believe so.

It'll require two JOINs, I think.

http://www.mediawiki.org/wiki/Categorylinks_table and http://www.mediawiki.org/wiki/Interwiki_table

And http://www.mediawiki.org/wiki/Page_table

MZMcBride, could you write it please?

Heh. I would if I could. :-)

;)

alnokta: There are some examples here that may help: http://en.wikipedia.org/wiki/User:MZMcBride/CSD

MZMcBride, i cannot write sql, i run it only. ;)

RSS: sunsecurity: Sun Alert 240866 Security Vulnerability in Solaris 10 NFS Remote Procedure Calls (RPCs) May Allow a Denial of Service (DoS) or Data Integrity Issues for Non-Global Zones <http://blogs.sun.com/security/entry/sun_alert_240866_security_vulnerability>

anyone volunteers to write the query?

Hmm... maybe.

Pathoschild looks.

alnokta: Which category?

Pathoschild, chemistry as a test

Pathoschild, i will need to run it across several ones to get several subjects covered

for the project: http://tinyurl.com/arwikiday

alnokta: I assume you want article namespace only?

Pathoschild, yes

This should probably work:

SELECT page_title FROM page LEFT JOIN langlinks ON page_id=ll_from LEFT JOIN categorylinks ON page_id=cl_from WHERE page_namespace=0 AND cl_to="Chemistry" AND ll_lang="ar" LIMIT 20;

Hm.

Actually, that gives you articles *with* a link to ar; getting articles without one will be a bit trickier.

cannot i just add ll_lang !="ar" ?

No, that will return (for example) "Neuroscience" because it has a link to fr, which is !="ar".

select page_title from page join categorylinks on page_id = ll_from and page_namespace = 0 and cl_to = 'Chemistry' where not exists ( select 1 from langlinks where ll_from = page_id and ll_lang = 'ar');

oh that is wrong

It's on the right track, though. How about this?

that http://rafb.net/p/g7sFJ814.html

SELECT page_title FROM page LEFT JOIN langlinks ON page_id=ll_from LEFT JOIN categorylinks ON page_id=cl_from WHERE page_namespace=0 AND cl_to="Chemistry" AND (SELECT 1 FROM langlinks WHERE ll_lang="ar" LIMIT 1) LIMIT 20;

Pathoschild nods.

thank you both! :)

Welcome. :)

Pathoschild: you're still looking for ones that do match in that query

Typo; just throw in an appropriate negation. :)

could it exclude redirects ?

add 'and not page_is_redirect = 1' somewhere

at the very end before the ;

Or "AND page_is_redirect=0", no?

well, yeah

or not page_is_redirect != 0 etc

:p

so which one to use patho's or carl's ?

makes no difference

Both work (I prefer mine because it's simpler).

for some reason I tend to assume page_is_redirect is either null or 1

which is why I write it that way

They're all 0 or 1 (I just checked). :)

yeah, the column is not null, but I never think of that

why that not working ? select concat(* [[:en:,page_title,]])

https://wiki.toolserver.org/view/%2Cpage_title%2C

you have to put the things you want to contat in quotes

concat('[[en:', page_title, ']]')

https://wiki.toolserver.org/view/'%2C_page_title%2C_'

beside the fields of corse :)

yes ignorance isn't an excuse ;) thx carl-m

make sure to add the final ') to what I wrote above

very good indeed http://ar.wikipedia.org/.../%D8%A7%D8%B3%D8%AA%D8%B9%D9%84%D8%A7%D9%85%D8%A7%D8%AA

nice %s ;)

;)

uninstalled hp's gcc distribution. turns out openssh was linked with gcc. now it won't start because libgcc.sl is missing ;)

-static-libgcc really needs to be the default

a static openssh is useful, yes :)

i don't want the entire thing static, just libgcc

@replag

Stwalkerster: s1: 3s [+0.00 sec/sec]; s2: 2s [+0.00 sec/sec]; s3: 1s [-0.16 sec/sec] (6h 32m 1s sample)

RSS: linuxkernel: 2.6.27-rc4-git5: stable snapshot <http://www.kernel.org/kdist/fragments/stable_snapshot.html>

_mary_kate_: hi

_mary_kate_: have you ever used boost.asio?

fale: lots of people in #boost have

_mary_kate_: thx :)

no one has complained about internal server error all day, does that mean they went away?

Duesentrieb: i patched mod_fastcgi again to ignore duplicate Status

yay. "ignore" by using the first or the second?

Duesentrieb votes for second

the second

or third, if there is one

hm, i'm actually not sure it's legal to do that though

mod_fastcgi's license only allows you to use it to implement standard protocols approved by the author, i.e. fastcgi

that doesn't sound dfsg-free

it's not

that's why it's in non-free :)

odd that apache doesn't have a free implementation instead

is mod_proxy_fcgi any good?

most people using fastcgi have probably moved to better webservers ;)

i don't see that one in the manual

found it by luck http://mproxyfcgi.sourceforge.net/

when I was looking for the mod_fastcgi license

seems dodgy

hmm, i don't see the point of it

why would you proxy _everything_ to php?

why would you make a module and require that it can onlybeused to implement one protocol?

so many questions

well, apache proxy has a seperate module for every protocol

mod_proxy_ftp, mod_proxy_http..

presumably they are freely licensed - that's the strange part

oh, mod_fastcgi. well it was written by the company that created fastcgi

maybe we should just change server and people using complicated .htaccess can live with it ;)

is server performance that much of a bottleneck?

for us in particular I mean

i'm more concerned about crappy mod_fastcgi

I know essentially nothing about webserver performance

hemlock has been pretty idle since we moved php to switchboard.. 10-20% cpu

that's great

switchboard is an alternative to mod-fastcgi?

no

it requires mod_fastcgi to work

what does it do then?

it takes php requests and sends them to a php process owned by the user who owns the script

and it can use one php process to handle several requests, instead of having to create a new php on every request

ah, so for every user there is a php process waiting?

not for every user, that would use an insane amount of ram. it creates processes on demand

_mary_kate_: heh: "mod_fcgid has a new process management strategy, which concentrates on reducing the number of fastcgi server, and kick out the corrupt fastcgi server as soon as possible. "

ah nice

i look at mod_fcgid but i couldn't find any way to use an external process with it

scratch that off then

I like the double-entendre aspect of their goal - I read is as "reduce the number of webservers running fastcgi"

heh