Re: [squid-users] Squid redirectors children doesn't shutdown/die with squid

From: Jorge Iván Burgos Aguilar <jorgeburgos_at_inaipyucatan.org.mx>
Date: Wed, 8 Sep 2010 10:03:27 -0500

Hi Amos,

See inline comments

2010/9/7 Amos Jeffries <squid3_at_treenet.co.nz>:
> All helpers MUST respond to every line of input they receive with a single
> line of output. Every non-response is a hung client request. They MAY
> respond with by shutting themselves down without output if they deem the
> input to be corrupt.
>
> Empty lines should not be sent to the url_rewriters. Are you sure this is
> not the end-of-line sent by squid for the lines you are reading in?

Well to be honest m not so sure, i have seen the code at
src/redirect.cc and cant confirm the behavior m seen, but if i put a
debug line logging all input send to the redirector script i find
thousands of \n (0x0a), a good line of arguments and then another
thousands of \n (0x0a), maybe as you say this is not send by squid but
maybe something related to the way python read the stdin, don't know
but obviously as of the moment m sending a \n per every line i found
that doesn't have the well-know syntax and seems to work...
====
          arguments = line.strip('.- \n').split(' ')

          # Validation
          if ( len(arguments) < 6 ):
             sys.stdout.write( "\n" )
             continue
====
The strange thing here and the thing that make ask for this behavior
is because if i don't send that "\n" and just ignore the redirector,
it does not work well, they start to die quickly and sometimes even
lag themselves, at the moment i'm considering that behavior some kind
of keep-alive related thing but not so sure so if someone here
enlighten me more about if m doing well or is a strange bug, it will
be very appreciated.

> perl spawned foreach into several more recent languages. Python being
> newer may have that or some equivalent.

OK, will see but as of the moment m thinking in the subprocess library.

Best Regards
Received on Wed Sep 08 2010 - 15:03:34 MDT

This archive was generated by hypermail 2.2.0 : Wed Sep 08 2010 - 12:00:03 MDT