help request: Based on the JAVA language, how do plug-ins make high availability

[trayliang]

Description

If the plug-in is written in JAVA, then when there is a problem with the JAVA program, when the JVM goes down, apisix will lose traffic. May I ask, is there a solution for this

Environment

• APISIX version (run apisix version):
• Operating system (run uname -a):
• OpenResty / Nginx version (run openresty -V or nginx -V):
• etcd version, if relevant (run curl http://127.0.0.1:9090/v1/server_info):
• APISIX Dashboard version, if relevant:
• Plugin runner version, for issues related to plugin runners:
• LuaRocks version, for installation issues (run luarocks --version):

[sunheyi6]

No, while it's not possible to completely avoid traffic loss, you can take measures to minimize downtime and restore the system quickly to reduce the impact on users. By implementing high availability, automated recovery, and fault tolerance mechanisms, you can ensure that the system can recover quickly when issues arise and minimize the duration of service interruptions. These approaches can help you effectively manage and mitigate the impact when Java applications fail, minimizing traffic loss as much as possible.

[trayliang]

thanks, Will the plug-in extension plan for high availability in the future

[hanqingwu]

maybe you can deploy multi apisix for a cluster , and put a LB in front of apisix cluster

[trayliang]

At present, we have completed the production of business expansion based on Java plug-in.
Our current deployment is clb -> apisix -> java-plugin to business server
Where apisix and java-plugin are deployed to the same container because they communicate with unix domain sockets,
Then we rationally analyze, Java-plugin is a java process, if the process is down, apisix will pull up the Java process again, but in the process of pulling up
Traffic loss occurs.

[hanqingwu]

OK, I get you point. So do you want apisix can retry mechanism when communicate timeout ?

[trayliang]

Retry for the C end is prone to overall timeout, I think can be solved from the internal, support plug-in high availability it

[trayliang]

Looking at the communication architecture between apisix and plug-ins, it is difficult to find a suitable theory to solve this problem

[trayliang]

But can you solve this problem from the apisix side, when the service call plug-in times out, redirect to other nodes, complete the channel