[squid-users] Survey on assertions: When the impossible happens

Alex Rousskov rousskov at measurement-factory.com
Tue Mar 1 17:37:52 UTC 2016


On 03/01/2016 03:55 AM, Eray Aslan wrote:
> On Mon, Feb 29, 2016 at 09:43:09AM -0700, Alex Rousskov wrote:
>> Q2: Your Squid is asserting every 5 minutes. There is no [working] Squid
>> version you can switch to. Your network topology does not allow you to
>> bypass Squid. Until the bug is fixed, would you prefer to see fewer
>> assertions in exchange for more memory leaks and an increased
>> probability of malformed/corrupted/misleading HTTP messages?

> False dichotomy.

Unfortunately, it is often a real one. In the real world, Squid is often
a single point of failure without good bypass options.


> Worst case:  Live through the outage, learn from it

Learning from failures is hardly the worst case. The worst cases in the
real world include innocent admins losing their jobs, kids exposed to
content they cannot unsee, etc., etc. It should not be that way, but it
sometimes is.


> and hopefully design your systems accordingly in the future.


These questions are exactly about "designing your systems" better! Squid
is a "system" itself, and if you think that there is always a way to
bypass Squid, then it should be easy for you to accept the same [false]
premise that there is always a way to bypass an assertion inside Squid.
Designing Squid to bypass internal failures is what options #2 and #3
are about. Same premise, same architectural principles, different zoom
level.


Cheers,

Alex.



More information about the squid-users mailing list