Re: [PATCH] Implement persistent id for proxies and servers

From: Willy Tarreau <>
Date: Tue, 4 Mar 2008 06:04:19 +0100

Hi Krzysztof,

On Fri, Feb 29, 2008 at 12:19:20PM +0100, Krzysztof Oledzki wrote:
> On Thu, 28 Feb 2008, Willy Tarreau wrote:
> >On Sat, Feb 23, 2008 at 01:22:41AM +0100, Krzysztof Oledzki wrote:
> >>>From be4c911d87a6ac800a1fb33828fa87b5fc2806a6 Mon Sep 17 00:00:00 2001
> >>From: Krzysztof Piotr Oledzki <>
> >>Date: Sat, 23 Feb 2008 01:19:10 +0100
> >>Subject: [PATCH] [MINOR] Implement persistent id for proxies and servers
> >>
> >>This patch adds a possibility to set a persistent id for a proxy/server.
> >>Now, even if some proxies/servers are inserted/deleted/moved, iids and
> >>sids can be still used reliable.
> >
> >I like the idea a lot. I have merged your patch (with a few minor typos
> >such as BACKED->BACKEND, or CVS->CSV).
> Right, sorry. :(
> >Right now, I'm a bit worried about the requirement to assign ids > 1000
> >and the fact that we don't check for conflicts (I know people with almost
> >500 instances).
> Hm. I assumed that 1000 should be fine as no one is supposed to have so
> many proxies.

you know the difference between what people are supposed to do and what you encounter in production :-)

> Obviously I was mistaken. We can bump this number (10K?) or
> write a documentation warning that with *really* big installations custom
> IDs should start with 10K. Of course checking for conflicts is still
> necessary, however it is better if they do not occur. ;)

yes, and if we check for conflicts, then we should be able to skip fixed numbers in automatic numbering. Let's say that a first version requires high values and that a later one will be able to automatically assign values avoiding conflicts.

> >But we'll be able to improve this later. I have already identified that
> >proxies should be dual-linked. It will then be easier to move instances
> >in the list, then automatically assign numbers to the unnumberred ones.
> Currently IDs are assigned to all instances and then overrried by a
> peristant ID. I did it intentionally not to mix up IDs of other instances
> if some get a persistent one.

I think this is an excellent idea.

Now I see how we would proceed to correctly assign the IDs :

best regards,
willy Received on 2008/03/04 06:04

This archive was generated by hypermail 2.2.0 : 2008/03/04 06:15 CET