TheLetterX has script problems

Is that the toolserver.org:81?

yes. except if you opt for this test, your normal URLs will use the new server

(which means if you use .htaccess, you should test it on :81 first)

RSS: linuxkernel: 2.6.27-rc5-git1: stable snapshot <http://www.kernel.org/kdist/fragments/stable_snapshot.html>

toolserver-l: 03River Tarnell * 10/: [Toolserver-l] new server testing, stage 2

3(created) [ST-47] STBotI mess up welcome message on user talk page; ST47's tools: STBotI; Bug <10https://jira.toolserver.org/browse/ST-47> (Kwan Ting Chan)

Hmm

_mary_kate_: -bash: /opt/zeus/webadmin-4.3r4/bin/convert_rewrite: No such file or directory

toolserver-l: 03Agony * 10/: Re: [Toolserver-l] help test a new web server

river seen lately?

_mary_kate_ == river

Werdna: thank you :)

heh

toolserver-l: 03SQL * 10/: Re: [Toolserver-l] help test a new web server

toolserver-l: 03SQL * 10/: Re: [Toolserver-l] help test a new web server

toolserver-l: 03MinuteElectron * 10/: Re: [Toolserver-l] help test a new web server

hi

toolserver-l: 03SQL * 10/: Re: [Toolserver-l] help test a new web server

[[ZWS]] 10https://wiki.toolserver.org/.../index.php?title=ZWS&diff=1364&oldid=1363&rcid=1715 * Bryan * (+11) (+on hemlock)

[[Tools outside Toolserver]] N 10https://wiki.toolserver.org/.../index.php?title=Tools_outside_Toolserver&rcid=1716 * Никола Смоленски * (+783) (New page: This is a list of tools that you might expect to find on the Toolserver, but are hosted somewhere else. * [http://vs.aka-online.de/wikipedia.html AKA's Wikipedia tools] ** GlobalWPSearch ...)

SQLDb: convert_rewrite is on hemlock

hi _mary_kate_

hello

Werdna wrestling with lighttpd

they seem not to have a such thing as boolean conditionals in their config files :/

$variable = "value" { opts... }

boolean

as in $HTTP["host"] == 'wiki.epstone.net' || $HTTP["host"] == 'testwiki.epstone.net' { opts.. }

_mary_kate_: I'm tempted to give up and just put the same config section twice with the separate conditions.

use a real web server ;)

or use ~= (a|b)

lighttpd is better than apache...

ew.

but that's better than putting the same section twice.

in my completely meaningless apachebench test, lighttpd managed only 15,000 req/s, while 44,000 req/s, while sjs web server managed 75,000 req/s

I don't mean that in a speed sense. I just like lighttpd better.

3(created) [TS-153] enable the new web server for my url's; Toolserver; Task <10https://jira.toolserver.org/browse/TS-153> (Luxo )

Luxo: you could just ask here ;)

:P

anyway, done

thx

_mary_kate_: where you come from? I thought USA, there is night :D

UK

Duesentrieb: if people ask for their stuff to be converted to zws, just add a proxypass in apache2.conf like the existing one for ~river

zws?

oh, the new webserver thingy?

kk

yerah

_mary_kate_: Proxy error: Error reading from remote server and now I don't get even that reply...

ok, now same error (Error reading from remote server), it just took ~three minutes to appear

why does CatScan take minutes to get the end of the HTML page across, while the CSV comes instantly?

something is not closing or flushing the connection as it should

is that the FCGI switchboard?

might be logging-related. though that would effect CVS too.

no clue, actually

Duesentrieb: Could you remove closed-zh-twwiki-p from toolserver.wiki? It seems to have been deleted, and it doesn't seem to have a replicated database on the toolserver.

am I the only one experiencing this problem, or do you see it as well?

LA2: i have seen it before, but inconsistently

Pathoschild: removing stuff by hand is always a bit dogy, because it'll resurface at some point. but if the db is no longer there, i guess it's ok

is the long result list built with string concatenation in a way that wastes far too much memory, and the process gets thrown out?

Thanks. :)

that could be a difference between HTML output and CSV

perhaps do a print for each line instead of building the entire page as one string?

Pathoschild: gone

LA2: i don't.

though php may be doing some buffering

but... no i don't think it does.

results often come "incrementally".

often 90% of the result list comes, and I'm waiting many minutes for the last rows

Duesentrieb: Whee, thanks. :)

LA2: sorry, i simply don't know why... and it's sadly very tricky to find out.

_mary_kate_: does the FCGI switchboard provide any statistics on how CatScan behaves, to aid Duesentrieb ?

statistics don't help.

oh, it always does

i'd need to disable all buffering, profile each step, and hope that doing that doesn't interfere with the cause of the problem.

i have timing stats in my own log. i suspect that might be a couase of the trouble. but as i said, that would affect CST too

are you ure it never happens for CSV?

not sure

do the output bytes pass through the FCGI switchboard?

or how does that work?

what is "xx_labswikimedia_p"?

(xx = lang)

Agony: en.labs.wikimedia.org

earlier test.wikipedia.org?

Agony: no, they are seperate tests

RSS: linuxkernel: 2.6.27-rc5-git2: stable snapshot <http://www.kernel.org/kdist/fragments/stable_snapshot.html>

ST47: you about?

Sure, sort of

Trying to figure out why PWP won't save

anyone here good with sql and screens?

toolserver-l: 03Magnus Manske * 10/: Re: [Toolserver-l] help test a new web server

hi guys

hi fale

hi SQLDb

hmm what would be the proper way to limit the execution time of a child process?

while reading stdout

alarms don't work with threads

oh of course select

ulimit?

that appears to be something shell related

i think you are right. dunno what it's using internally - because that's what you would want to be using too, i suppose

though starting child processes via a shell may be an option. depending on what you are doing

does wikipedia.de run on any of toolservers?

donated

Wikimedia Deutschland donated some

saper: you probably misunderstood me. i wonder if the wikipedia.de site runs on any of servers from toolserver group or if it is hosted anywhere else

netname: DOMAINFACTORY-20070312

netname: DOMAINFACTORY-20070312 descr: Infrastructure

wikipedia.de and wikimedia.de are within 80.67.25.0/24

tools.wikimedia.de (and other toolservers) are on 91.198.174.0/24

that's netname: WIKIMEDIA-EU-NET

so, they are most probably hosted completely separately