Re^2: Is anyone at perlmonks aware that...

Bypassing DNS round-robin in my wgets...

In the last few minutes, when I get content from 216.92.34.251, 100% of the time it is zero bytes (bad).

Also, at this time, when I get content from 209.197.123.153 and 66.39.54.27, the content is a typical 68K page 100% of the time. (good)

Is there an admin who can look at md5sums or diffs of the programs which generate content at the bad server vs the good servers?

UPDATE: I am only referring to a GET of the "/" URI.

Comment on Re^2: Is anyone at perlmonks aware that...

Replies are listed 'Best First'.
Re^3: Is anyone at perlmonks aware that... (zero bytes) by tye (Sage) on Jun 02, 2011 at 18:02 UTC
I don't think an md5sum between mod_perl 1 and mod_perl2 would be very informative. :) Thanks for reporting the IP addresses involved. Too often, people report something like "perlmonks.org is broken but perlmonks.com works" which is much less helpful (since I can't query their system to see which of the three IPs associated with perlmonks.org is currently being used by their browser nor which of the exact same 3 IPs associated with perlmonks.com is currently being used by their browser). The zero-byte results that I have been able to reproduce (for certain nodes when viewed by certain people but not on the problematic 216 IP) have been due to core dumps. Meanwhile, the logs on the 216 IP have been completely unhelpful. I'm tempted to just request that the 216 be removed from the DNS for now, though I'm hesitant because it could lead to more over-all site slowness. So far, it seems that the problem shows up for a fairly short while and then disappears. That makes me think it might be worth requesting an Apache config change so mod_perl children don't live quite so long. I'm pretty sure pair.com monitors the web services (and I suspect the blank pages might trigger pair.com to restart Apache which could explain why the problem goes away fairly quickly). I should try to find if that monitoring data is actually available to me and might contain any useful information in diagnosing the problem. - tye	[reply]

Replies are listed 'Best First'.

Re^3: Is anyone at perlmonks aware that... (zero bytes)
by tye (Sage) on Jun 02, 2011 at 18:02 UTC

I don't think an md5sum between mod_perl 1 and mod_perl2 would be very informative. :)

Thanks for reporting the IP addresses involved. Too often, people report something like "perlmonks.org is broken but perlmonks.com works" which is much less helpful (since I can't query their system to see which of the three IPs associated with perlmonks.org is currently being used by their browser nor which of the exact same 3 IPs associated with perlmonks.com is currently being used by their browser).

The zero-byte results that I have been able to reproduce (for certain nodes when viewed by certain people but not on the problematic 216 IP) have been due to core dumps. Meanwhile, the logs on the 216 IP have been completely unhelpful. I'm tempted to just request that the 216 be removed from the DNS for now, though I'm hesitant because it could lead to more over-all site slowness.

So far, it seems that the problem shows up for a fairly short while and then disappears. That makes me think it might be worth requesting an Apache config change so mod_perl children don't live quite so long.

I'm pretty sure pair.com monitors the web services (and I suspect the blank pages might trigger pair.com to restart Apache which could explain why the problem goes away fairly quickly). I should try to find if that monitoring data is actually available to me and might contain any useful information in diagnosing the problem.

- tye

[reply]