[Wikitech-l] slowness today

classic Classic list List threaded Threaded
32 messages Options
12
Reply | Threaded
Open this post in threaded view
|

[Wikitech-l] slowness today

Domas Mituzas
We had some slowness today, where after some temporary issue cluster ended
up spending >50% of it's time working on Spanish wikipedia template. We
nuked the template (es.wikipedia site was also turned off at some moment),
so now it is all up and running.

The template was used for adding a single line of text, though it took 20
seconds to render. ;-)

We'll probably have to do something about that.

Cheers for now,
Domas
_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Carlos-27
Domas Mituzas wrote:

> We had some slowness today, where after some temporary issue cluster ended
> up spending >50% of it's time working on Spanish wikipedia template. We
> nuked the template (es.wikipedia site was also turned off at some moment),
> so now it is all up and running.
>
> The template was used for adding a single line of text, though it took 20
> seconds to render. ;-)
>
> We'll probably have to do something about that.
>
> Cheers for now,
> Domas

Can we do temporary edits to the template to make it semi-functional
(only a couple of switchs), or it can't be touched until you finish
investigating?

--


_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l

Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Platonides
> Can we do temporary edits to the template to make it semi-functional
> (only a couple of switchs), or it can't be touched until you finish
> investigating?

It's 5 days since this reminding, a week since the slowness. We're
missing the
funcionality it provided. So, is there any reason blocking from
restoring it?
(if so, why hasn't been resolved yet? :D)

There was a 'policy' of users don't needing to worry about server issues
('if there aren't enough servers, we will add more'), but here there is
a collision.
As any edit to the template will affect a lot of pages, i'm also
reluctant of doing 'lighter' versions, as changes+reverts would probably
overload even more. Who should we send versions for approval?


_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Jens Frank
On Tue, Jan 16, 2007 at 07:47:40PM +0100, Platonides wrote:
>
> There was a 'policy' of users don't needing to worry about server issues
> ('if there aren't enough servers, we will add more'), but here there is
> a collision.

That was before wiki markup became a real programming language. Now it's
possible to write templates that require a substantial amount of time to
render (20s for the interproject template you're asking about).

As every programmer, a wiki template programmer has to take care of the
efficiency of his code. Adding servers will not improve the rendering
time of slow templates.

Regards,

        jeluf

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Gregory Maxwell
On 1/16/07, Jens Frank <[hidden email]> wrote:
> That was before wiki markup became a real programming language. Now it's
> possible to write templates that require a substantial amount of time to
> render (20s for the interproject template you're asking about).
>
> As every programmer, a wiki template programmer has to take care of the
> efficiency of his code. Adding servers will not improve the rendering
> time of slow templates.

Since there is still no (real) looping constructs, all wikitext
execution should still be roughly linearly proportional to input size,
or am I missing something?

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Brion Vibber
In reply to this post by Platonides
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Platonides wrote:

>> Can we do temporary edits to the template to make it semi-functional
>> (only a couple of switchs), or it can't be touched until you finish
>> investigating?
>
> It's 5 days since this reminding, a week since the slowness. We're
> missing the
> funcionality it provided. So, is there any reason blocking from
> restoring it?
> (if so, why hasn't been resolved yet? :D)
>
> There was a 'policy' of users don't needing to worry about server issues
> ('if there aren't enough servers, we will add more'), but here there is
> a collision.

Jesus Christ!

I can't say words without having some God-given "policy" read into them.

I made a general recommendation not to go running around saying THE SKY
IS FALLING THE SKY IS FALLING about templates BASED ON SUPPOSITION AND
PARANOIA.

That doesn't mean that AN ACTUAL PROBLEM, WHEN DISCOVERED, SHOULD BE
IGNORED.

WHEN THERE IS AN ACTUAL, REAL, MEASURABLE PROBLEM, THEN IT MATTERS.

IN THIS CASE THERE IS AN ACTUAL, REAL, MEASURABLE PROBLEM.

UNTIL THAT GETS FIXED UP ON THE SOFTWARE END, THE PARTICULAR TEMPLATE
WHICH IS AN ACTUAL, REAL, MEASURABLE PROBLEM, TAKING 10-20 SECONDS TO
RENDER, MUST NOT BE WIDELY USED.

Try to use some common sense, people.

There's no "policy" here. Just *sense*.

- -- brion vibber (brion @ pobox.com)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFFrTZ5wRnhpk1wk44RAm20AKDOOw0muLgX/Zc6PBA+ghuyajbV2wCg2UM/
pYPy5UjcDDZB6g2gkTSiuO8=
=zNMT
-----END PGP SIGNATURE-----

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Carlos-27
In reply to this post by Gregory Maxwell
Gregory Maxwell wrote:

> On 1/16/07, Jens Frank <[hidden email]> wrote:
>
>>That was before wiki markup became a real programming language. Now it's
>>possible to write templates that require a substantial amount of time to
>>render (20s for the interproject template you're asking about).
>>
>>As every programmer, a wiki template programmer has to take care of the
>>efficiency of his code. Adding servers will not improve the rendering
>>time of slow templates.
>
> Since there is still no (real) looping constructs, all wikitext
> execution should still be roughly linearly proportional to input size,
> or am I missing something?

If I recall correctly (people in this list know better) template
inclusion and ParserFunction operations are done by repeteadly
generating new strings. I suppose that as the strings get larger these
operations get slower... maybe that's the cause of the slowness (maybe not).

--



_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Rob Church
In reply to this post by Brion Vibber
On 16/01/07, Brion Vibber <[hidden email]> wrote:
> I can't say words without having some God-given "policy" read into them.

The English Wikipedia took comments you and I made about being
sensible and not worrying *too much* in a previous thread, and turned
it into a guideline;
http://en.wikipedia.org/wiki/Wikipedia:Don%27t_worry_about_performance.

I assume, or hope, that these people still know when to suspend
guidelines. I think most of them got it.

> That doesn't mean that AN ACTUAL PROBLEM, WHEN DISCOVERED, SHOULD BE
> IGNORED.

To quote: "Adopt common sense, of course. If it's plain something
could cause drastic problems, hold fire and check... -- (me)" -- this
qualifies as causing problems...

...also, ""Policy" shouldn't really concern itself with server load
except in the most extreme of cases; -- (brion)" -- and I imagine this
qualifies as an extreme case.

> Try to use some common sense, people.

Fervent agreement!

> There's no "policy" here. Just *sense*.

If someone tells me that the English Wikipedia's horrible habit of
taking things as written, as a suicide pact...has crept over to our
end...I'm not sure if I'd be able to grok that.


Rob Church

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Platonides
In reply to this post by Brion Vibber
Brion Vibber escribió:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Platonides wrote:
>>> Can we do temporary edits to the template to make it semi-functional
>>> (only a couple of switchs), or it can't be touched until you finish
>>> investigating?
>> It's 5 days since this reminding, a week since the slowness. We're
>> missing the
>> funcionality it provided. So, is there any reason blocking from
>> restoring it?
>> (if so, why hasn't been resolved yet? :D)
>>
>> There was a 'policy' of users don't needing to worry about server issues
>> ('if there aren't enough servers, we will add more'), but here there is
>> a collision.
>
> Jesus Christ!
>
> I can't say words without having some God-given "policy" read into them.
I quote me: "here there is a collision" between this general guidelines
and this case.


> I made a general recommendation not to go running around saying THE SKY
> IS FALLING THE SKY IS FALLING about templates BASED ON SUPPOSITION AND
> PARANOIA.
>
> That doesn't mean that AN ACTUAL PROBLEM, WHEN DISCOVERED, SHOULD BE
> IGNORED.

> WHEN THERE IS AN ACTUAL, REAL, MEASURABLE PROBLEM, THEN IT MATTERS.
> IN THIS CASE THERE IS AN ACTUAL, REAL, MEASURABLE PROBLEM.
Damn, if i though that it didn't matter, i would have just reverted it,
or even duplicated, not asking for the opinion of THOSE WHO KNOW.


> UNTIL THAT GETS FIXED UP ON THE SOFTWARE END,
Of course, this is what we want: it being solved :D


Brion...
> THE PARTICULAR TEMPLATE
> WHICH IS AN ACTUAL, REAL, MEASURABLE PROBLEM, TAKING 10-20 SECONDS TO
> RENDER, MUST NOT BE WIDELY USED.

...and JeLuF
"As every programmer, a wiki template programmer has to take care of the
efficiency of his code."

I refered to this before.
How is the efficiency of the templates measured? Apart of just trying it
and waiting to see if servers go down :P


Diclaimer:
I didn't code it, i didn't even watched its source until today. I don't
say it to be a good style. It isn't easy to completely understand. I
have even found what seems to be bugs.
So don't eat me!

I know there's a problem. What is being done about it? What can we do to
help with it? How do we do it?


_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Carlos-27
In reply to this post by Brion Vibber
Brion Vibber wrote:

> Platonides wrote:
>
>>>Can we do temporary edits to the template to make it semi-functional
>>>(only a couple of switchs), or it can't be touched until you finish
>>>investigating?
>>
>>It's 5 days since this reminding, a week since the slowness. We're
>>missing the
>>funcionality it provided. So, is there any reason blocking from
>>restoring it?
>>(if so, why hasn't been resolved yet? :D)
>>
>>There was a 'policy' of users don't needing to worry about server issues
>>('if there aren't enough servers, we will add more'), but here there is
>>a collision.
>
> Jesus Christ!
>
> I can't say words without having some God-given "policy" read into them.
>
> I made a general recommendation not to go running around saying THE SKY
> IS FALLING THE SKY IS FALLING about templates BASED ON SUPPOSITION AND
> PARANOIA.

Now we are confused, running in circles crashing into each other and
taking decisions based on suppositions, because...

> That doesn't mean that AN ACTUAL PROBLEM, WHEN DISCOVERED, SHOULD BE
> IGNORED.
>
> WHEN THERE IS AN ACTUAL, REAL, MEASURABLE PROBLEM, THEN IT MATTERS.
>
> IN THIS CASE THERE IS AN ACTUAL, REAL, MEASURABLE PROBLEM.
>
> UNTIL THAT GETS FIXED UP ON THE SOFTWARE END, THE PARTICULAR TEMPLATE
> WHICH IS AN ACTUAL, REAL, MEASURABLE PROBLEM, TAKING 10-20 SECONDS TO
> RENDER, MUST NOT BE WIDELY USED.

...what we want to know is very simple:

a) are you going to change mediawiki so it works, or
b) should we forget it and make a new one (and in that case, do you know
what construction exactly should be avoided, or should we guess and try
at Special:Expandtemplate), or
c) you don't yet know?
d) [other option I didn't think about]

And people are already implementing different solutions that they think
are best, multiplying different templates, while others are waiting for
that template to be functional again, and lots of articles meanwhile
lose their interproject links... *because nobody knows what to expect*.

So... could any of you please give us some information about this
specific problem, so people at eswiki can make an informed decision?

> Try to use some common sense, people.
>
> There's no "policy" here. Just *sense*.

'Policy' was between quotes and Platonides is not a native speaker.
Probably he meant something like 'rule' (probably not, but I understood
it that way), and he explicitly said that it doesn't apply here.
es:WP:PBF :-P

Greetings.
--



_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Brion Vibber
In reply to this post by Platonides
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Platonides wrote:
> How is the efficiency of the templates measured? Apart of just trying it
> and waiting to see if servers go down :P

This is very easy:

1) Stick it on a page
2) Hit "preview"
3) react in shock and horror at how long it took to render
4) run it several times and compare against lack of template to confirm
that, yes, it is simply adding that template which causes 100-fold
increase in rendering time


> I know there's a problem. What is being done about it? What can we do to
> help with it? How do we do it?

Between other things, someone sooner or later will get a chance to poke
at the software to try to fix the performance problem.

- -- brion vibber (brion @ pobox.com)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFFrUMIwRnhpk1wk44RApf8AKDW8Qx6nUrDaRcfYcYVMZ5xEA6ABwCggzMy
xucJ8VDM3juaMeXxsqAbFDo=
=8Ir8
-----END PGP SIGNATURE-----

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Brion Vibber
In reply to this post by Carlos-27
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Carlos wrote:
> ...what we want to know is very simple:
>
> a) are you going to change mediawiki so it works, or

Yes.

> b) should we forget it and make a new one (and in that case, do you know
> what construction exactly should be avoided, or should we guess and try
> at Special:Expandtemplate), or

Do that anyway; the template's horribly ugly!

And I'm sorry for my harsh words.

- -- brion vibber (brion @ pobox.com / brion @ wikimedia.org)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFFrURTwRnhpk1wk44RAjdiAJ4w2JW+XVxGTrX89RjfI7uPyqNpewCfQLiQ
d6bD7i/EUwJF853DZ+92VbU=
=4Eti
-----END PGP SIGNATURE-----

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Carlos-27
Brion Vibber wrote:

> Carlos wrote:
>
>>...what we want to know is very simple:
>>
>>a) are you going to change mediawiki so it works, or
>
>
> Yes.
>
>
>>b) should we forget it and make a new one (and in that case, do you know
>>what construction exactly should be avoided, or should we guess and try
>>at Special:Expandtemplate), or
>
>
> Do that anyway; the template's horribly ugly!

Thanks! :)
--



_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

David Gerard-2
In reply to this post by Brion Vibber
On 16/01/07, Brion Vibber <[hidden email]> wrote:

> Jesus Christ!
> I can't say words without having some God-given "policy" read into them.
> I made a general recommendation not to go running around saying THE SKY
> IS FALLING THE SKY IS FALLING about templates BASED ON SUPPOSITION AND
> PARANOIA.
> Try to use some common sense, people.
> There's no "policy" here. Just *sense*.


Fortunately, this has been written up as policy: [[m:Don't be dense]].
Phew! Writing it up as policy should fix the problem!


- d.

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Brion Vibber
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

David Gerard wrote:

> On 16/01/07, Brion Vibber wrote:
>> Jesus Christ!
>> I can't say words without having some God-given "policy" read into them.
>> I made a general recommendation not to go running around saying THE SKY
>> IS FALLING THE SKY IS FALLING about templates BASED ON SUPPOSITION AND
>> PARANOIA.
>> Try to use some common sense, people.
>> There's no "policy" here. Just *sense*.
>
>
> Fortunately, this has been written up as policy: [[m:Don't be dense]].
> Phew! Writing it up as policy should fix the problem!

*cries*

:)

- -- brion vibber (brion @ pobox.com)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFFrhUgwRnhpk1wk44RAj9hAJ9nyJV6tr8Ys+6iXKomUYicSFR9BwCeNGcy
wsx3PUrC35Mj/QOkQfO95JM=
=TWah
-----END PGP SIGNATURE-----

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Rob Church
In reply to this post by David Gerard-2
On 17/01/07, David Gerard <[hidden email]> wrote:
> Fortunately, this has been written up as policy: [[m:Don't be dense]].
> Phew! Writing it up as policy should fix the problem!

We also now have the [[Wikipedia:Everything Brion says is gospel]]
policy, which means we also have [[Wikipedia:Brion is on crack]] and
[[Wikipedia:Wikipedia sucks]] policies now.


Rob Church

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Jay Ashworth-2
On Wed, Jan 17, 2007 at 04:16:02PM +0000, Rob Church wrote:
> We also now have the [[Wikipedia:Everything Brion says is gospel]]
> policy, which means we also have [[Wikipedia:Brion is on crack]] and

ROTFL.  Really.

Cheers,
-- jra
--
Jay R. Ashworth                                                [hidden email]
Designer                          Baylink                             RFC 2100
Ashworth & Associates        The Things I Think                        '87 e24
St Petersburg FL USA      http://baylink.pitas.com             +1 727 647 1274

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Domas Mituzas
In reply to this post by Platonides
>
> There was a 'policy' of users don't needing to worry about server issues
> ('if there aren't enough servers, we will add more'), but here there is
> a collision.


Two things are infinite: the universe and human stupidity; and I'm not sure
about the universe. -- Albert Einstein
_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

Rob Church
On 18/01/07, Domas Mituzas <[hidden email]> wrote:
> Two things are infinite: the universe and human stupidity; and I'm not sure
> about the universe. -- Albert Einstein

:D


Rob Church

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: [Wikitech-l] slowness today

David Gerard-2
In reply to this post by Domas Mituzas
On 18/01/07, Domas Mituzas <[hidden email]> wrote:

> > There was a 'policy' of users don't needing to worry about server issues
> > ('if there aren't enough servers, we will add more'), but here there is
> > a collision.

> Two things are infinite: the universe and human stupidity; and I'm not sure
> about the universe. -- Albert Einstein


http://uncyclopedia.org/wiki/Stupidity_Reactor


- d.

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://lists.wikimedia.org/mailman/listinfo/wikitech-l
12