Discarding old VCL

[Poul-Henning Kamp]

In message <AANLkTim4N80jQrHbwUMhpo2Cf-uGjmzwWT9PPFVPHnC1 at mail.gmail.com>, Mich
ael Lenaghan writes:
>>>Any idea where that ref count is coming from?
>>
>> Some worker thread, which must be stuck somewhere.
>>
>> What is the oldest non-zero refcount you have seen ?
>
>Hello.
>
>It looks like the ref counts do clear eventually--but it can take
>several days. On 6/26, for example, there was a config from 6/22 that
>still had a ref count of 1. By 6/27 it dropped to zero. As of today,
>6/28, the oldest config is from 6/25. So perhaps it's strange--or at
>the very least unexpected--but not really a problem.

I don't think there is a problem per se.

The idle worker threads hold a reference to the last VCL they used
when they sleep, so until you have exercised all your threads during
peak-load, you may have old VCL's linger around.

Not optimal obviously, but you would have to load a lot of VCL's
for it to become a problem.

Poul-Henning

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.