Mailing List Archive

>
> If I'm right about the cause of this problem, this patch ought
> to do something about it. Could someone give it a whirl?

Nope, sorry.

ooo.lanl.gov - - [15/Nov/1995:09:20:42 -0700] "NULL" 200 0


This was with a SSI calling a CGI. The CGI script just
did a sleep for longer than TIMEOUT.

>
> If I'm right about the cause of this problem, this patch ought
> to do something about it. Could someone give it a whirl?

Can someone else give it a try.

I tinkered with some stuff yesterday, and can't be 100% sure that
my tinkering didn't disable Rob's patch.


try a html file calling a cgi script which just does a long
sleep.

rob

D'oh! These are *sub*-requests, not internal redirects! Sorry about
that... as Bullwinkle sez, this time for sure:

*** http_main.c~ Fri Oct 27 16:37:34 1995
--- http_main.c Wed Nov 15 14:37:59 1995
***************
*** 246,252 ****
* if it times out...
*/

! log_transaction(timeout_req);

pfclose (timeout_req->connection->pool,
timeout_req->connection->client);
--- 246,260 ----
* if it times out...
*/

! request_rec *log_req = timeout_req;
!
! while (log_req->main || log_req->prev) {
! /* Get back to original request... */
! if (log_req->main) log_req = log_req->main;
! else log_req = log_req->prev;
! }
!
! log_transaction(log_req);

pfclose (timeout_req->connection->pool,
timeout_req->connection->client);

> D'oh! These are *sub*-requests, not internal redirects! Sorry about
> that... as Bullwinkle sez, this time for sure:

Okey dokey, that one fixed it, thanks.

Could the bug have caused a meltdown or was it a coincidence that
I saw the NULL at the time of the meltdown?

rob

I can't imagine how this particular bug could have caused a server
meltdown... it causes the wrong request_rec to get logged, but I can't
see how it could create *multiple* log file entries from a single bogus
transaction, and you said you had thousands of 'em.

rst

> I can't imagine how this particular bug could have caused a server
> meltdown... it causes the wrong request_rec to get logged, but I can't
> see how it could create *multiple* log file entries from a single bogus
> transaction, and you said you had thousands of 'em.

I see the same here. There are often several tens of them; nicely groupwd, and
all with a corrospoinging out of memory error in the error log. However I have
not had a server meltdown on exactly that.... that comes later.. when memory
gets *really* thight.

Dw.


--
http://me-www.jrc.it/~dirkx http://ewse.ceo.org
DWvGulik@Dialis.xs4all.nl Dirk.vanGulik@jrc.it
+39 332 78 1322 +39 332 78 9549
fax +39 332 78 9185

ISEI/ESBA; The Center For Earth Observation
Joint Research Centre of the European Communities, Ispra, Italy