[squid-users] cache_peer round robin

[Amos Jeffries]

On 18/05/2023 6:58 pm, Alexeyяр Gruzdov wrote:
> Hello guys !
>
>
> *1. Question Number 1: *
>
> A day ago I met with strange things of squid working.
>
> I have twenty of cache_peer and these all works in round-robin mode
> that means each new requests will operated by new cache_peer.
>
> For example the five  from these twenty cache_peer is down. But others 
> is OK and live ( this is fifteet)
>
> When I run my tests script for check - that I can see the first around 
> 10 - 30 seconds all looks OK and each new requests is going over next 
> one cache_peer.
> But after these 10 - 30 seconds the round-robin mechanism stops to 
> operate and all new requests are going over one cache_peer (look like 
> this is FIRSTUP_PARENT but not the ROUNDROBIN_PARENT )

Which mechanism is your Squid using to detect the peer recovery?
Have you perhapse used "default" selection option on any of the peers?

What cache_peer lines are you using exactly? (IP/port can be obscured)



>
>
> *2. Question Number 2:*
>
> Is there some mechanism to use some external script based on the 
> events of squid?
> for example - squid generates a log message like
>     "Detected DEAD Parent: proxy_42662"    ( that means some one 
> cache_peer unavailable ) and will be great to listens these message 
> and doing some actions ( the easiest its notification )


Yes, but not for that particular event.

HTH
Amos