#394 Transifex http errors
Closed: Fixed None Opened 16 years ago by mmcgrath.

Transifex http port is unresponsive at times. Its been difficult to track down. Externally if you run:

nmap -p 80 translate.fedoraproject.org

in a loop, it will eventually come up closed. Anyone that has a moment to take a look and troubleshoot please do.


I think this happens in Damned Lies as well.

I'm bumping priority since it affects public pages of ours and contributions.

<nod> I think I'm finally on to something here.

Unlike the https (port 443) farm, the port 80 farm actually does more then just a TCP connect, it requests a page. If that page doesn't come back in 1 second it flags that server as dead.

Long term fix: Ensure the main landing page for transifex loads real quick.

Short term fix: I've created some mod_rewrite rules that might help:

{{{
RewriteCond %{REMOTE_ADDR} 10.8.32.12
RewriteRule ^/$ http://localhost/ [P,L]

RewriteCond %{REMOTE_ADDR} 10.8.32.13
RewriteRule ^/$ http://localhost/ [P,L]
}}}

Basically what I'm saying here is, if the remote host is one of our balancers, load from localhost. This should be a very quick operation. I'm going to leave the ticket open for now, lets see how this goes over the next week or so.

Update: Nice, I haven't seen this problem for the past month of so :)

I haven't either. I'm going to close this ticket.

Noticed again today, several times.

To reproduce, visit https://translate.fedoraproject.org/ and notice that the CSS aren't loaded half of the times.

This should be completely fixed as of a while ago. if it is not please re-open.

Login to comment on this ticket.

Metadata