Re: Announcing branch: cachemgr-refactoring

From: Amos Jeffries <squid3_at_treenet.co.nz>
Date: Fri, 04 Jul 2008 22:38:03 +1200

Kinkie wrote:
> On Fri, Jul 4, 2008 at 11:42 AM, Henrik Nordstrom
> <henrik_at_henriknordstrom.net> wrote:
>> On fre, 2008-07-04 at 10:24 +0200, Kinkie wrote:
>>
>>> First thing, whatI already mentioned on -dev.
>>> cachemgr is over HTTP, might as well show that up. This is a very
>>> simple task but it'll make life MUCH easier for external tools'
>>> developers.
>> Yes, and somewhing we all agreed on I think..
>
> Yes.
>
>>> That's another very interesting possibility, but I haven't really
>>> thought of it. The data cachemgr displays are semi-structured but the
>>> structures differ quie a lot between one view and the next. If a
>>> common ground can be found, it'd be interesting to follow it.
>> Just forget the existing cachemgr displays if looking for structure.
>> Instead start from the SNMP data as Adrian said, and I also said earlier
>> in the cachemgr discussion. Trying to convert the existing cachemgr
>> displays into something more structured simply isn't worth the effort
>> when there already is a well structured view of pretty much the same
>> data in the SNMP interface.
>
> I agree. Maybe even more structure and data can be shown than what's
> in SNMP, and that's exactly the kind of things I'd like to think
> about.
>
>> The existing cachemgr interface is really only meant as a debug and
>> diagnostics tool. Data collection is best done using SNMP.
>
>> Enabling an XMLRPC view of the SNMP data is an interesting topic, and
>> would enable a whole range of new tools access to the Squid performance
>> counters and status indicators.
>>
>> If there is things missing in the SNMP MIB these should be added there,
>> instead of trying to invent new structures for this data.
>
> How would everyone feel about changing the cachemgr (possibly breaking
> backwards compatibility) so that it can render better an high-level
> view of the dataspace, while leaving detailed views (possibly even
> more detailed than now) to snmp, xmlrpc and/or structured (but
> user-unfriendly) html / plaintext displays?

I don't see much use in it. As Henrik said, better to make cachemgr read
data from the SNMP interface.

IMO, the better long-term on data collection and display would be to
create a simpler OO way for new structures/views to be added to SNMP.

Amos

-- 
Please use Squid 2.7.STABLE3 or 3.0.STABLE7
Received on Fri Jul 04 2008 - 10:38:04 MDT

This archive was generated by hypermail 2.2.0 : Fri Jul 04 2008 - 12:00:03 MDT