Hello, Am Montag 03 September 2007 23:01 schrieb Sean Whitton:

Hello,

To confirm, are you suggesting we edit our scripts to read this file and put in "sorry, server is down" if that is the case?

You should read the file and parse the first line, that not beginns with "#". Then your tool knows, what's the status of cluster is. Is it "OK", you have to do nothing more, if it's "INFO" and "WARN", it's your decison if you display something. Is it "ERRO" or "DOWN" you must ("must" like in rfc 2119 [1]) give out the text. But that is the better decision anyway, because your tools did not work because the cluster is down :). If you add additional informations in your tools, is your desicion of corse.

This only applies to tools using sql too, right?

Of corse :). If you need no mysql, information about mysql is useless, isn't it? Perhaps there can be a general-information-file in future (for warning about planing down-time of the hole cluster) - but not yet.

I am personally against an enforcable rule (but would definately follow it if imposed).

The enforcable is ONLY for ERRO and DOWN. And if you have good reasons to use it not, then use it not (but we send all users to you ;)).

BTW: Erwin asked in chat, if there is a way to read the files without using http. Yes, there is. The files can be found in /var/www/

Thanks,

Sean

Sincerly, DaB.

[1] http://tools.ietf.org/rfcmarkup?doc=2119

You could quite easily load the file when they choose a particular db, or read all the DBs statuses and hide their entries on the box or reject any wikis from them from being selected.

-Matt

Date: Tue, 4 Sep 2007 00:08:57 +0200 To: toolserver-l@lists.wikimedia.org From: platonides@gmail.com Subject: Re: [Toolserver-l] SQL-S2 down and new status-files

DaB.:

...

You should read the file and parse the first line, that not beginns with "#". Then your tool knows, what's the status of cluster is. Is it "OK", you have to do nothing more, if it's "INFO" and "WARN", it's your decison if you display something. Is it "ERRO" or "DOWN" you must ("must" like in rfc 2119 [1]) give out the text. But that is the better decision anyway, because your tools did not work because the cluster is down :).

What about tools not using a fixed db (depends on user choose)? Also, being only a line i'd make it the first one.

---

Toolserver-l mailing list Toolserver-l@lists.wikimedia.org http://lists.wikimedia.org/mailman/listinfo/toolserver-l

_________________________________________________________________ Explore the seven wonders of the world http://search.msn.com/results.aspx?q=7+wonders+world&mkt=en-US&form=...

Hello,

Could someone kindly remind me which box user databases are on (1, 2 or 3?)?

Thanks,

Sean

On 05/09/07, Byrial Jensen byrial@vip.cybercity.dk wrote:

DaB. wrote:

...

To inform our tool-user better in future, I created some files:

http://tools.wikimedia.de/status_s1 (same for s2 and s3).

That accounts for connections to sql-s1, sql-s2 and sql-s3.

What about connections to sql to access the toolserver database? Shouldn't there be a status file for that?

Best regards, Byrial

---

Toolserver-l mailing list Toolserver-l@lists.wikimedia.org http://lists.wikimedia.org/mailman/listinfo/toolserver-l

Sean Whitton:

Could someone kindly remind me which box user databases are on (1, 2 or 3?)?

user databases are on "sql". you shouldn't rely on them being on a particular cluster, since they might move to another one (after zedler crashed, i moved user dbs from zedler to vandale).

- river.

Sean Whitton:

Okay, how can I tell that this server is down using this new system?

i created another status file, "status_sql". (it's actually a symlink to the server currently holding user databases).

- river.