After playing with static clustering a few more days, I still ran
headlong into a wall with this.
I decided to go back to multicast, and try BackupManager instead of
DeltaManager, and it now seems to be restarting correctly, as long as my
timeout was high enough for the 15+ second AbstractMapReplication init.
Overall it takes about 65 seconds for server startup compared to 15 sec
for static deltaManager and 30 sec for multicast deltaManager. But the
important part is that it eventually does start the server while under
load, and session failover seems to work as expected.
Unfortunately using backupManager is not my preferred solution, but it
may be the only option until I can figure out why DeltaManager is broken
in our environment. I would still love to hear if anyone has any
suggestions as to why backup manager is working but delta manager is broken.
Thank you,
-Anthony
On 3/25/2016 10:28 AM, Anthony Sturchio wrote:
Thanks for the response. I apologize if I'm using the incorrect
terminology here, as this is one of the smaller "hats" that I wear at
work.
Basically, coldfusion runs on top of a (possibly customized) tomcat
backend. Each instance has its own server.xml and web.xml. As best I
can figure, when starting CF, it starts tomcat in the JRE, and if
applicable, starts deltamanager to handle the clustering business.
After the sessions are replicated, the servlet (coldfusion server)
starts up with the connector ports specified in server.xml. This all
happens relatively smoothly in testing without any appreciable load.
However when the box is added back into our webserver farm (via
hardware load balancer), or if I create artificial load against the
box using apache jmeter, session replication is not successful, and
the coldfusion server never loads. I don't get any errors written to
screen or log, nor does kill -3 give me anything, but as per ps aux,
the java process is still running. The replication ports and Catalina
connector ports are open, however the overall server shutdown port is
not open, if that offers any clues.
Web pages are being served via apache httpd and mod_jk and AJP/1.3.
Thank you,
-Anthony
On 3/25/2016 10:09 AM, Christopher Schultz wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Anthony,
On 3/25/16 9:41 AM, Anthony Sturchio wrote:
We recently updated our Coldfusion 10 server to update 18, which
moved us up to Tomcat 7.0.64, and we are now experiencing issues
when restarting a CF instance. I understand that this is not a
coldfuison forum, but since CF10 is based on top of a tomcat back
end, which is where the issue appears to be, I figured I would ask
here. As best I can figure, it appears that while under moderate
load, DeltaManager sessions dont replicate, and the servlet hangs
and never fully starts up. Without any load, the CF instance
(servlet) starts up perfectly fine without issue.
Can you clarify what it means for a servlet that has not yet started
up to be "under load"?
- -chris
-----BEGIN PGP SIGNATURE-----
Comment: GPGTools -http://gpgtools.org
Comment: Using GnuPG with Thunderbird -http://www.enigmail.net/
iEYEARECAAYFAlb1RoIACgkQ9CaO5/Lv0PBpxgCggKFVpE/HM++CSwsGw73r6Yni
UHYAn2dcnX/FomVD19Tz+TjEe1cMi/Zd
=ruIF
-----END PGP SIGNATURE-----
---------------------------------------------------------------------
To unsubscribe, e-mail:users-unsubscr...@tomcat.apache.org
For additional commands, e-mail:users-h...@tomcat.apache.org
