Re: [squid-users] Delay Pools bug?

On 28/02/2014 2:15 p.m., Dan Charlesworth wrote:
> Hi folks
>
> We have two Squid boxes (one with 2.7, one with 3.4.3), which have almost identical delay pool configurations, but the 3.4.3 box is limiting the speed to around half of what the 2.7 one is; around 32KB/s when it should be 64KB/s
>

We have a strange situation in Squid-3 with delay pools. Both these seem
to be happening:

overcounting
 http://bugs.squid-cache.org/show_bug.cgi?id=3536

undercounting
 http://bugs.squid-cache.org/show_bug.cgi?id=522

Delay pools is a very weak and buggy form of QoS. In general it is
better to use the tcp_outgoing_tos/_mark features and use the far better
system QoS controls.

> Relevant parts of the configs:
>
> 2.7 Config
> —————
> acl 1_cmps src “/path/to/iplists/policy_1"
>
> delay_pools 1
> delay_class 1 2
> http_access allow 1_cmps
> delay_access 1 allow 1_cmps
> delay_parameters 1 -1/-1 65536/65536
> delay_initial_bucket_level 100
>
> 3.4.3 Config
> —————
> external_acl_type bandwidth_type ttl=300 children-startup=2 children-idle=1 children-max=10 %URI %EXT_LOG bandwidth_ext_acl.py
>
> delay_pools 1
> delay_class 1 2
> acl bandwidth_ext_acl external bandwidth_type 512
> http_access allow bandwidth_ext_acl
> delay_access 1 allow bandwidth_ext_acl
> delay_parameters 1 -1/-1 65536/65536
> delay_initial_bucket_level 100
>
> I’m not able to reveal any of the specific external ACL code, but that shouldn’t make any difference should it?
>

No that wont matter for this.

Amos
Received on Fri Feb 28 2014 - 07:34:21 MST