| time |
nick |
message |
12:38 |
<jduty> |
amp rocks |
12:39 |
<jduty> |
all the slowdowns we saw with lots of outstanding messages went away when using amp |
12:40 |
<rajdavies> |
amp? |
12:41 |
<jduty> |
oh.. sorry, the amq message store |
12:41 |
<rajdavies> |
cool! |
12:41 |
<jduty> |
is very fast and light weight |
12:41 |
<rajdavies> |
its just got a bit faster too in trunk |
12:42 |
<jduty> |
we saw a huge perf decrease when using journaled jdbc persistence as messages built up in subscriptions |
12:42 |
<jduty> |
did the same test yesterday with almost 200K outstanding messages... almost no decrease |
12:42 |
<rajdavies> |
nice to hear it works |
15:27 |
<srjk> |
hey |
15:27 |
<srjk> |
when I send out recipt tags with a send command to a queue I am subscribed to, I only get the recipt for the message. I dont get the message ? |
15:27 |
<srjk> |
is that right ? |
15:29 |
<chirino_m> |
srjk: right |
15:30 |
<srjk> |
so when coding a client, is it better to not send the recipt tag for send commands (Ones I am subscribed to ) ? |
15:30 |
<srjk> |
or not look for a message ? |
15:30 |
<srjk> |
or is there no best practise ? |
15:42 |
<srjk> |
i'm going with takign the message on a subscribed to desst. on a send as the receipt |
15:42 |
<srjk> |
seems reasonable |
15:44 |
<chirino_m> |
srjk: you should only requests recipts in case you want confirmation that the message will not be lost. |
15:44 |
<chirino_m> |
but if you are subscribed to those same messages, it's not really needed since you get confirmation when you get your own message back I guess. |
15:47 |
<srjk> |
chirino_m: cool thanks |
19:39 |
<ACM> |
hi |
19:40 |
<ACM> |
anyone know a good reason why activemq could run out of 10240 file descriptors? |
20:04 |
<ACM> |
kinda quiet here ? |
20:05 |
<chirino_m> |
ACM: that's alot |
20:05 |
<ACM> |
yeah |
20:05 |
<chirino_m> |
ACM: you on linux ? |
20:05 |
<ACM> |
yes, with a recent 2.6 kernel |
20:05 |
<chirino_m> |
run: lsof | grep pid |
20:05 |
<chirino_m> |
where the pid is the pid of the activemq broker |
20:06 |
<chirino_m> |
should give you a listing of all the open files |
20:06 |
<chirino_m> |
let me know what the most of them are. |
20:06 |
<chirino_m> |
are they files or sockets? |
20:06 |
<ACM> |
its now stable at 106 - 109 |
20:06 |
<ACM> |
in total |
20:06 |
<ACM> |
mostly references to the jars |
20:06 |
<ACM> |
but when I had 9000+ earlier, it were mostly sockets |
20:07 |
<chirino_m> |
ah |
20:07 |
<chirino_m> |
well, it could be you were getting DOSed |
20:07 |
<chirino_m> |
what version? |
20:07 |
<ACM> |
5.1 |
20:07 |
<chirino_m> |
hum |
20:07 |
<ACM> |
but I doubt we were dossed |
20:07 |
<ACM> |
Let me check the web-stats |
20:08 |
<chirino_m> |
well next time it happens take a stack dump of the broker |
20:08 |
<ACM> |
no real huge peaks |
20:08 |
<chirino_m> |
the stack dump will show what those connections were doing. |
20:09 |
<ACM> |
you mean with jhat? |
20:09 |
<ACM> |
euh, jheap |
20:09 |
<chirino_m> |
jstack |
20:09 |
<ACM> |
hmm, now I'm mixing up commands :) |
20:09 |
<ACM> |
ah, I see it |
20:10 |
<ACM> |
thanks for your input so far |
20:11 |
<chirino_m> |
k |
