XMPP Service Operators - 2018-04-16

@ping trashserver.net

@ping shreddox.ei

nuron: Ping failed (remote-server-not-found): Server-to-server connection failed: DNS resolution failed

@ping shreddox.eu

nuron: Pong from shreddox.eu in 3.253 seconds

nuron: Ping failed (remote-server-not-found): Server-to-server connection failed: connection-timeout

@ping trashserver.net

Its down. First i thought it was my usual problem with gajim and wicd again.

nuron: Ping failed (remote-server-not-found): Server-to-server connection failed: connection-timeout

> Its down. First i thought it was my usual problem with gajim and wicd again. Yes, trashserver.net is down. There is a problem with mam_muc module. I think Thomas try to fix it..

Have the same problem with prosody after I updated the modules

@ping trashserver.net

nuron: Pong from trashserver.net in 4.023 seconds

Its up again

What problem, do you know?

The mam_muc module doesn't work...

Maybe it's this: <https://hg.prosody.im/prosody-modules/rev/39994c6bb314>?

I will send it Thomas..

Thanks

He just States there is 'a problem' https://social.tchncs.de/@trashserver

And earlier ranted about prosody and MySQL 😃

Of course, because I have texted him that mam muc doesn't work ;)

There are no problems, there are opportunities.

Challenges 😃

We'll see

Martin: Isn't challenges the thing we had 10 years ago, while we've now realized this still sounds way too negative? I might be mixing things up.

Holger: maybe I'm not up to date with these euphemisms

The trashserver admin sees an opportunity to test https://github.com/ortuman/jackal it seems :-)

Lets call everything celebration and try to solve all problems with party and dope 😃

\o/

> The trashserver admin sees an opportunity to test https://github.com/ortuman/jackal it seems :-) It seems he is eveb thinking about getting involved in development

.even

Yes, he is

MySQL

It's-bad-soo-bad-Postgres-soooo-much-bettta!

Maybe but since everything supports it while most doesn't support the other

I can't tell if that was sarcasm or not, but it really is… if you have a datbase that you've started with that you're forced to use, fair enough. If you're starting fresh and use MySQL you're just asking to lose data.

I'll just keep swearing in dead languages at it whenever InnoDB tables just decide to corrupt 'emselves

Maranda: Oh I thought this was the rant you were about to make :-) I'm not into SQL engines and don't care.

I say that, InnoDB is actually a fairly nice storage backend in a lot of ways. But MySQL's nonsense still makes it not worth using.

Mostly innodb won't corruptp itself to my knowledge, but it will decide that you obviously don't know what you're doing with your query and it should make a lot of assumptions and corrupt your data

SamWhited, *my* query? 😎

Everybody's query!

SamWhited: You lost me, InnoDB won't corrupt itself but what?

The queries aren't *mine* I'd need to ask the web junkies who wrote 'em. But not that it matters much

PostGRE didn't corrupt with the equivalent queries

SamWhited: You can get it to corrupt data with queries that are bogus in some way? Or with valid queries InnoDB incorrectly considers bogus?

Holger: Sorry, I'm just ranting. As an example, mysql in its default mode will happily decide that because you accidentally queried for where intcolumn = '1_whatever' that you *obviously* meant hat to be "where intcolumn = 1" and it will select that and insert data even though your query was a mistake and you were trying to query against a string column or something

Holger: with invalid queries that shouldn't run

It also ignores constraints in a lot of cases; it's trivial to make non-nullable columns contain an ull.

Ah, I see.

It's worth noting that you can turn all this off and that mysql people will swear that this makes it fine… but naturally you have to know that and go to more effort to have any sort of data consistency.

SamWhited apparently though saving caches in MySQL "is not a thing" while it causes no issues in pgsql

Also, the utf-8 type isn't actually wide enough to encode utf-8… so they have another utf-8-for-real-this-time type that does it right… the whole thing really is a joke.

Maranda: saving caches? I'm not sure what you meant by that, sounds bad though.

We do that :-)

I have wasted a lot of time trying to make things work on mysql… so I'm a bit bitter now :)

With InnoDB.

(Saving/restoring it's main buffer/cache thing on restart, I mean.)

ah, gotcha

I had Orchard storing page caches in MySQL and it was corruption all over until I figured it was that

The same wouldn't do a thing in PG

Interesting; I don't know how innodb stores pages, I'd be curious to know why that lead to corruption.

It seems to work just fine for as. As usual with anecdotal evidence ...

For us even.

(Or maybe we're talking about different features.)

I don't care about the database (engine), as long as it's supported by the stuff I need to use and doesn't cause hassles.

This is not how flame wars work.

:P

One of the ways of making something not cause hassle is making sure you use the right thing for the right job

On the xmpp server I proudly use flat files for a reason.

MySQL in general will cause hassle though… I don't think there's a job it's uniquely suited for. If you want speed, don't use a relational database. If you want data consistency, use a different relational database.

MattJ, are you trying to citate Scotty on me :P?

The only reason to use mysql is because you already have a database that uses it and you're stuck with it.

SamWhited, or that the stuff you need to use or that people need you to use requires it?

Yah, fair

Although I'd also consider using other tools at that point; requiring mysql gives me no confidence in your <insert project> or that it won't lose my data.

But that's obviously not always an option.

SamWhited: We've been using hints to fix MySQL's optimizer choosing wrong indices for weird corner cases. I hear that's frowned upon in the Postgres universe, Postgres should just always get it right and if it doesn't, post a bug report. Sounds really good to me, except if there's a realistic chance I could run into the situation where I have to post a bug report and wait for a fix.

SamWhited, I just give you an example, all mainstream web analytics apps (mostly php) don't or badly support pgsql

Of course there may really be enough magic in Postgres that this won't happen in practice, I have no idea.

Apart from that I've seen long lists of reasons to prefer Postgres and not much vice-versa, yes.

Holger: I have never had it get it wrong, but if it does I don't think it's impossible to direct it to use a different index, they just ask that you also file a bug report.

Holger, well posting a bug report about MySQL doesn't work much better either...

That being said, the data for most of the projects I've worked on wasn't that complex, so it may have trouble in places. I'm sure it's not right 100% of the time.

SamWhited: As I understood it there's no explicit directive to do that (on purpose). But there may be tricks to do it of course.

I'm not sure from how much time in practice MySQL workbench migration from Postgres is just plainly broken.

Maranda: My point was that MySQL offers you a workaround for the time until the bug is fixed.

Holger: yah, don't listen to me, I thought there was but I don't see it now. Might have been cockroach DB feature that I'm mixing up or something.

Good thing I'm not a DB expert at all

:P

Does innodb support multicolumn indexes? I've purged my mysql knowledge after leaving a previous job…

Oh yes it does.

SQLite does :-)

https://upload.lightwitch.org:5280/share/335b8d23-0aca-4a81-966d-0683a9055137/scottySays.jpg

MattJ, ^

SamWhited: But I have no real doubt that Postgres is probably the superior product. MySQL's success story seems to boil down to "everyone uses it" and that's basically how I got to know it better than Postgres as well.

Also now it's Oracle's says it all.

Holger: yah, it was the same with me the first time I used a database, "well this seems to be what everyone is using, must be the best"

It's also "fast", which is why a lot of people like it. But it's speed comes from taking shortcuts that lead to data inconsistency like not checking constraints so you probably want something else if that's why you're using it…

Maranda: there's always MariaDB in that case (the non-oracle fork which is actually what I assume people are using for the most part when they say 'mysql' these days)

and if it's speed me thinks Pgsql10 is faster anyways now

When I say we used "MySQL" at a previous job, we were actually on MariaDB I think. I shouldn't conflate the two, but I forget sometimes.

err, no, that's not true either, we had both. Gosh I'm forgetting all this.

SamWhited, except afair MariaDB has no Windows release so I'm stuck with Oracle's :P

\\o

o//

SamWhited: True in my case, I meant MariaDB when I said MySQL.

mam_muc error mod_mam_muc should be loaded only on a MUC component, not normal hosts i found this in the prosody logs...

what does it mean?

what it says.

that you're trying to load it somewhere which is not a MUC component.

but how can i fix that? i have no idea

put it in modules_enabled = {} pertaining to (aka below) whatever *Component "..." "muc"* section?

and not somewhere else or globally or somesuch

I'm sorry my English is not very well. I will try to understand..

I have mam_muc in the modules_enabled list and in the virtual host part... Which one should I delete?

nuron: both

Okay

Component "conference.example.org" "muc" modules_enabled = { "mam_muc", }

An then?

Put it under the "Component" line

Yes, there is it. Between only name, restrict_room_creation and max_history_messages

Component "conference.shreddox.eu" "muc" name = "Shreddox Muc's" restrict_room_creation = false max_history_messages = 100 modules_enabled = { "mam_muc" }

thats the part

That one is correct, you can remove the other one.

okay, i've done so. restart prosody now?

Yup, that should fix it :)

i'll try

just a second

no don't restart

so, prosody is up again

lets have a look

(that's pointless I think)

.

brb

why no restart?

there is no mam yet

no mucmam

1:1 mam should work

nuron: what version of Prosody is this?

0.10

Ah, it's not fully compatible with 0.10. It should work partially, but some clients won't be able to detect it I think

i have updated the modules today a few hours ago and since this update mam_muc dont work..

See the note under "Compatibility" here: https://modules.prosody.im/mod_mam_muc.html

but i have used it three month ago and everything works fine

with prosody 0.10 as well

I'm not sure when this happened; someone in the prosody room might be able to give you more information

i will check the link first

https://haste.tchncs.de/yujumudolo.pas That was also in the logs

and very often this log: conference.shreddox.eu:mam_muc debug <history xmlns='http://jabber.org/protocol/muc' since='2018-04-16T18:33:05.221Z'/>

dont know what to do :/

Apr 16 20:24:26 s2sin76a5ad0 debug Incoming s2s connection Apr 16 20:24:26 s2sin76a5ad0 debug Incoming s2s received <stream:stream version='1.0' xmlns='http://etherx.jabber.org/streams'> Apr 16 20:24:26 mod_s2s debug sending: <?xml version='1.0'?> Apr 16 20:24:26 mod_s2s debug sending: <stream:stream xmlns:stream='http://etherx.jabber.org/streams' version='1.0' id='28cb310f-b8fb-4398-8527-8c8eabeb4913' xmlns:db='jabber:server:dialback' xmlns='jabber:server'> Apr 16 20:24:26 s2sin76a5ad0 warn No 'to' on stream header from unknown host means we can't offer any features Apr 16 20:24:26 mod_s2s debug Sending stream features: <stream:features/> Apr 16 20:24:26 mod_s2s debug sending: <stream:features> Apr 16 20:24:26 s2sin76a5ad0 debug Received[s2sin_unauthed]: <verify id='167c0bd9' xmlns='jabber:server:dialback' to='lightwitch.org' from='jit.si'>ca4847b1d2223b51899d27cfbce08a44153544b49bfc4ce45ddacfe8b58b8d83</verify>

wowow

that's ancient

@version jit.si

Maranda: jit.si is running Openfire version 3.10.3 on Linux 4.4.0-22-generic (amd64) - Java 1.8.0_91

nuron: I'd ask in the prosody room, they're more likely to be able to help you.

xmpp:prosody@conference.prosody.im?join

im allready there ;)

Hmm I almost thought it was jabberd, but I vaguely remember OF doing it to verify dialback streams (or some such)