Good morning, i need your help please
In a Production system an application comprises TomEE and the core application, and it is distributed among 4 nodes or servers
Many many times the URL for node1 its not responding, sometimes there is a delay of 2 mitutes to load,sometimes i get a 404 error, so the temporal solution is to restart tomEE almost daily, but this is not a good practice because this URl receives thousands https requests from internet.
behind those 4th nodes it is a load balancer, so these questions arise:
1 when and why should i restart tomEE? how frequently?
2 in Unix/Linux itself is a way to see for each node the number of request being received from the load balancer? so i can know if LB is sending the request in a balanced way to each node?
I am new to Unix and i have no idea about web servers applications, if u have any url to get documented
I appreciate your help in advanced
Be warned that bumping posts are against the rules on this forum and moderators might sanction you for it.
Which load balancer (LB) are you using??
However, I would be likely to compare the 4 configurations (conf/server.xml) to explore any differences seen. TomEE still uses normal Tomcat Clustering and that needs checking out.
I would be inclined to guess that the LB is not overloading node1, on the contrary, the LB is continuing to send requests to node1 after that node has failed (for some reason). There is a fault (hardware, software or configuration) with node1.
The two minute delay you see to load the page is, in fact, the timeout before the request is successfully delivered to a surviving node. If everything is configured properly, it shouldn't take this long. Furthermore, once the failed node is kicked out of the cluster (until you issue the command for it to rejoin), the LB should know that node1 is dead and no longer send requests to it. You then run on three nodes after failover and all is fine and dandy.
So your problem is that failover is not occurring in an orderly fashion. I would confirm whether node1 is actually joining the cluster at all or you have a cluster of three plus a stand-alone node.
The way it should work is that the four cluster members should know immediately if one of them fails. The failed node gets kicked out. The surviving nodes relay this state to the LB which no longer sends requests to the failed node.
Good afternoon, Sorry i apologized for bumped questions, i will take into account
the data ive got so far is this; LB is F5, method used:least connection, persistence method:source adress location, i dont understand this terminology but it was given by our Networking ADministrator
Thanks you mery much for your help and feedback given
--- Post updated at 12:17 PM ---
one more issue: The company that developed the application made a comparison among these 4 confogurations and mention and confirmedthat are the same configuration, but they can not find what is happenning.
so i come up with this questions.
by means if any command how can i know if a node is kicked out of a cluster or how is it rejoined?