Closed Bug 299577 Opened 19 years ago Closed 19 years ago

searches for bugs that contain a word (not a string) timeout

Categories

(bugzilla.mozilla.org :: General, defect)

defect
Not set
normal

Tracking

()

RESOLVED FIXED

People

(Reporter: nelson, Assigned: justdave)

References

()

Details

In short: since the server move, bugzilla doesn't seem to be able to search 
for matches on a single word any more.

A search for all open bugs (unconfirmed, new, assigned, reopened} 
in all products, where a comment "contains all the words/strings"  md5
finds 136 bugs in less than 10 seconds.  But it finds bugs where md5
is a substring inside the middle of a bunch of uuencoded stuff, or in the
string CRAMMD5, which I don't want.  

So I change the search to be where a comment "contains all the words" md5
and the search times out.  I see "bugzilla is pondering your query" for 
about 15 seconds, and then the throbber stops, leaving me looking at that 
statement.  

I've tried all the following variants of that search:
1. a comment "contains any of the words" md5:  timeout
2. a comment "contains the string" md5  result, same as for "all the 
   words/strings" md5, the same 136 bugs
3. searching only in products { core, firefox, mozilla app suite, TBird }:
   timeout
Has this changed any since this last week's upgrade?
This bug is still reproducible today, using the steps described above.
I find it hard to believe that a search could get worse as a result of the server move.  We moved to a newer version of MySQL with much more efficient query parsing and index processing as part of that move.  It may be something else that's triggering it and the timing was just coincidental.

Can you give me the URL that's in your URL bar when one of these timeouts happens?  I tried a couple of your examples here, and they all give me results within 10 or 15 seconds.
I agree that the timing may have been merely coincidental.  
I noticed it immediately after the server move, while the bugzilla pages all
asked people to be on the lookout for problems, and to report them.
OTOH, there were some other things that changed as a result of the server 
move (e.g. a different version of interdiff, IIRC), so it could be more than
a mere coincidence.

Here are some URLs that timeout for me today:

https://bugzilla.mozilla.org/buglist.cgi?query_format=advanced&short_desc_type=allwordssubstr&short_desc=&long_desc_type=anywords&long_desc=md5&bug_file_loc_type=allwordssubstr&bug_file_loc=&status_whiteboard_type=allwordssubstr&status_whiteboard=&keywords_type=allwords&keywords=&emailassigned_to1=1&emailtype1=exact&email1=&emailassigned_to2=1&emailreporter2=1&emailqa_contact2=1&emailtype2=exact&email2=&bugidtype=include&bug_id=&votes=&chfieldfrom=&chfieldto=Now&chfieldvalue=&cmdtype=doit&order=Reuse+same+sort+as+last+time&field0-0-0=noop&type0-0-0=noop&value0-0-0=

https://bugzilla.mozilla.org/buglist.cgi?query_format=advanced&short_desc_type=allwordssubstr&short_desc=&product=Core&product=Firefox&product=Mozilla+Application+Suite&product=Thunderbird&long_desc_type=anywords&long_desc=md5&bug_file_loc_type=allwordssubstr&bug_file_loc=&status_whiteboard_type=allwordssubstr&status_whiteboard=&keywords_type=allwords&keywords=&emailassigned_to1=1&emailtype1=exact&email1=&emailassigned_to2=1&emailreporter2=1&emailqa_contact2=1&emailtype2=exact&email2=&bugidtype=include&bug_id=&votes=&chfieldfrom=&chfieldto=Now&chfieldvalue=&cmdtype=doit&order=Reuse+same+sort+as+last+time&field0-0-0=noop&type0-0-0=noop&value0-0-0=

If you moved via mysqldump, it could be that key distribution stats aren't up-to-date.  We should probably be running ANALYZE TABLE once a week anyway, but it might be worth running it now on the tables in the Bugzilla database to see if that fixes the problem.
(In reply to comment #5)
> We should probably be running ANALYZE TABLE once a week anyway,

OK, we now have a cron job on mrdb01 running as part of cron.weekly that does this.

> but it might be worth running it now on the tables in the Bugzilla database to
> see if that fixes the problem.

Done.  It didn't help.

However, I did discover that the send timeout in apache on recluse is set to 120 seconds (it was 300 seconds on mecha), so what do you know, it was related to the server move (just webserver, not database).  I just bumped it up to 300, and now, although said queries still take 2 minutes 30 seconds to run, they actually complete and give you results now instead of timing out.
Status: NEW → RESOLVED
Closed: 19 years ago
Resolution: --- → FIXED
Component: Bugzilla: Other b.m.o Issues → General
Product: mozilla.org → bugzilla.mozilla.org
You need to log in before you can comment on or make changes to this bug.