I have to correct my assertion. My preliminary tests were done long time
ago and Oxford is actually compiling from source code. There is no
2.7.STABLE9 version in the SL5 repository. Still I could upgrade without
problems a random machine using the rpm I found on the RH site. Although
unsupported it has all the standard features.
http://people.redhat.com/jskala/squid/
Oxford on the other hand has compiled the official source code with the
standard options. This, while easy enough, requires the manual creation
of the squid user and the setup of logrotate in an installation from
scratch.
I opened a savannah bug where I also attached the last experience I had
when investigating why the squid user wasn't configured without nologin
at most sites using the frontier rpm. I repeated the same experience
with the RH rpm and didn't have any problem with that.
http://savannah.cern.ch/bugs/?77447
I'm not sure where we are going with this because SL people need to be
involved to have an offcial rpm in the OS repository and am not sure
this will happen. Still if we don't complain they think there is no problem.
cheers
alessandra
On 25/01/11 20:07, Alessandra Forti wrote:
>
> > I do remember being told a reason that the standard squid wasn't
> good enough. I've forgotten though...
>
> my preliminary tests show there is no reason. On top of it Oxford is
> using the OS squid in production. The use of cvmfs is
> an additional reason to get away from frontier squid and be on a more
> reliable ground.
>
> Today at the dteam meeting this was discussed again and it was agreed
> I'd open a savannah ticket that could be used as reference.
>
> I will add your complaints about cvmfs. Mine about cvmfs would have
> been just a forecast (forecasts are never listened to., it seems
> people need to get drench before admitting it's raining) but since
> you've felt it on your skin it'll help to give it more weight.
>
> cheers
> alessandra
>
>>> I believe squids aren't very good on virtual machines because
>>> latency is
>>> important and thats something you lose with virtualization. At the
>>> Tier
>>> 1 we currently have 6 squid boxes, (doing 3 different things) but we
>>> hope that over time we can merge these into one large pool.
>>>
>> With CVMFS we ought to have 2 squids really I guess.
>>
>> Chris
>>
>>> Alastair
>>>
>>>
>>> On 24 Jan 2011, at 17:30, Alessandra Forti wrote:
>>>
>>>> The frontier squid is a mess personally if I can avoid using it in the
>>>> future I will. I showed the developers twice step by step how
>>>> difficult it was compared to the OS squid. The first time, after I
>>>> also attended their meeting, they removed customize.sh which overrides
>>>> the local configuration from the startup script. My point was more
>>>> general though, do sites with squid caches services really need a
>>>> frontier rpm? I think they talk about it but never reached a
>>>> conclusion.
>>>>
>>>> cheers
>>>> alessanda
>>>>
>>>> On 24/01/11 16:07, Stephen Jones wrote:
>>>>> We were on 7.3 here, and I also broke things while upgrading. It's
>>>>> working now, I think. It seems quite of lot of merging is needed to
>>>>> bring the new config in line with the old one (e.g. paths for caches
>>>>> and log files etc. are all different).
>>>>>
>>>>> Steve
>>>>>
>>>>> Christopher J.Walker wrote:
>>>>>> But I've managed to break things in the process of upgrading :-(.
>>>>>>
>>>>>> We are in downtime now for power work - but will try and fix it
>>>>>> when we
>>>>>> are back up.
>>>>>>
|