#musicbrainz-devel

w/e, it works.

well, rika is throttled to 50kbps

for the entire thing

that would probably do it.

except for stuff within the MB firewall, but

the notion being that even if someone tries to abuse it it can't be more than an insignificant drop compared to our real traffic

I assume it's blocked from search engines too?

luckily things like updating from git.musicbrainz.org are nice and fast, and you don't have to do CPAN stuff because I have it set up to copy those from a different version

yes

robots.txt is hardcoded for everyone

er, except navap actually

Still?

because he was playing with some search engine stuff

well, I haven't bothered to revert it, so

Oh, right I've got a file it's reading from. Well it's set to restrict all anyway

yeah

Does nginx look for config files within the MB source tree, or are the nginx files within admin/nginx copied out to etc/nginx, or something else entirely?

they're copied out, generally

possibly symlinked though

Well I made a change to MeB's nginx rewrites. I was wondering if that will take affect the next time ruaok does a git pull onto production or whether I need to remind him to copy the new file out

hm

no idea about metabrainz

do you know which server that's on? I can go see how it's set up

I don't

hm

pino, evidently

cool, it's symlinked in

Oh, I knew that.

so it should be fine as soon as it gets pulled

pino shows up in the git log

ah

aww, no fabric for that

I guess I'd need to VPN it anyway

I currently change the namespace in DBDefs when I want to "clear" memcached. Is there a more appropriate way to do it?

I don't know that there's a nicely-packaged way

What happens to a namespace once I change away from it? Does it eventually get discarded?

it should, yes

Then I'll just keep changing namespaces when I need to. It's really easy to do, no fancy script to remember.

yeah

I was figuring I'd just write a script to stick in your home directory and run :)

maybe even stick it into run.sh

so it just clears the cache if you kill/restart run.sh

What's run.sh?

I guess you predate that

it's just a script to start the server

in your case it'd be a bit weird since you'd need two

I just use the line you gave me many many moons ago

yeah, that's pretty much all that's in run.sh

When I need to I ctrl-c the server, press up, enter :)

navap: for your main install, you can now run clear-memcached.sh in your home directory

if you want it for your second install, copy that script and change musicbrainz-server to musicbrainz-server2, as you may imagine

navap: I've also copied over run.sh; I didn't include the memcached-clearing in that though

I hope you didn't write dump.pl just for this

I did, actually, but I've also wanted the same thing

and the other thing it does, which is just printing out everything that's in the cache

it's annoying to do by hand because, as you can see in the script, you have to do a 'stats items' and then a 'stats cachedump' on each slab

I also wanted the same thing on a previous project I worked on

because memcached is everywhere :P

basically: that script will continue to be useful for quiiiiite a while, I suspect :)

also you could use that script to dump anything that's a prefix, so e.g. all wikidocs pages

hm, when I run it I get Can't locate Cache/Memcached.pm

oh

hm

you probably need to cpanm it

because that's only in the carton local directory, not anywhere dump.pl can see it

alternatively do some bullshit to make it look in the musicbrainz-server local dir

PERL5OPT='-Mlocal::lib=/home/navap/musicbrainz-server/local' should do it, I guess, but that's gross

hello!

Is it possible to stick that line in clear-memcached.sh?

Moin warp!

navap: heh, actually, I have a different solution

navap: should work now

What did you do?

ran it with carton

it runs 'carton exec -- perl ../dump.pl' now :P

Ah

since there's no actual restriction that you have to run stuff that's within the directory, you just have to run it *from* the directory

Cool, it now clears memcached when I run run.sh from within my repos

what's fun is that while that'll destroy sessions it won't log people out who have selected the 'remember me' checkbox :)

which is a neat trick

Im guessing that problem Spacey had last night is he is using a 32bit VM instead of a 64 bit, and 32bit VM can only memory map to 4GB so cannot map all indexes

heh

so his needs to swap to nfio by using ?init=nfio parameter or swap to 64bit jvm

what's nfio?

(and I'll pass that along next time I see him, also)

nfio = new file io

ah

i.e regular files

navap: musicbrainz-server and metabrainz-server updated; nginx on pino restarted.

ruaok: I'm contributing some hopefully-useful blathering to the doc you sent out earlier :)

ijabz: that sounds quite plausible on Spacey

sweet.

I also chose a much nicer color this time, much as I love fuchsia :)

thanks. :)

oh yeah, perl6. :)

haha

yup

ruaok assuming the search server crash last night wasn't at the same time as a load of new indexes ?

ijabz: lets see

how many hours ago was that?

maybe 10pm your time?

approx, yes

they swapped in shortly after midnight UTC

ianmcorvidae: where this new doc ?

ijabz: https://docs.google.com/document/d/1bVll8H6MJjA...

so they went wrong before that then, oh well I have to take kids to school biab

midnight UTC is not 10pm your time is it?

k, tt

no its 11pm i think

ianmcorvidae: please feel free to add more points to the lang pros and cons

ianmcorvidae: on the general trend in that doc.

I can see that we might want more than one language.

but I want to prevent language proliferation. we already have three for the various bits of MB.

I would argue that we should not add *any* more languages.

the question still remains -- are we looking for a replacement for perl, or one language to rule them all unless we really *can't* do what we want with whatever we choose?

I think one language is not really feasible, nor really desirable.

so, not one to rule them all.

but in general I want to establish a policy that tells us when we should use what.

as per your suggestion maybe we need to define common tasks we need these languages to carry out for us.

and then define a language per task group/area/whatever

perhaps

I think that's a good way to get a large number of languages

what do you think the goal should be?

not sure.

(Midnight UTC is 1am UK DST, not 11pm, BTW.)

uk: i knew I should've looked it up. I can never get that one right.

I think the easiest thing to define it as is "what do we want to replace perl with", but I'm not sure about our ultimate goal/order-of-operations

Plus or minus, all the same. :)

dyslexia sucks. :(

e.g. the data access layer, whatever it is, will probably come first; that may merit a different language choice than the web interface; don't know if the webservice will be separate, etc.

sounds like we're headed towards a lib for the DAL.