Page MenuHomePhabricator

Move WLE map to new server
Closed, ResolvedPublic


The map lives on a Bahnhof server and should be moved to the new Glesys server.

Event Timeline

Sebastian_Berlin-WMSE changed the task status from Open to In Progress.Aug 19 2022, 10:24 AM
Sebastian_Berlin-WMSE claimed this task.
Sebastian_Berlin-WMSE created this task.

I managed to get it up and running, I think. I had to change server/server-build.gradle to get JWBF to work.

The map is running on the sites server now. The Apache site file looks something like:

<VirtualHost *:80>
        DocumentRoot /srv/wle-map/wle-2020-sweden-map/client/src/main/webapp/
        ProxyPass /api/ http://localhost:8080/api/
        ProxyPassReverse /api/ http://localhost:8080/api/

The server is currently crunching data and building up an index. The size of it on the new server is a the moment 19 MB compared to 137 MB on the old server. I'll have a look at moving the index to the proper directory. It looks like the path may be hard coded. now points to the new server. I set up a cron job that starts the server each reboot. Log is not being saved because quickly gets quite big. We could do something with logrotate if we want to.

The server takes about a quarter of the CPU and almost all when its starting up. Adding a file with org.gradle.jvmargs=-Xmx1000m makes sure that it doesn't stay at max CPU.

Lokal_Profil added subscribers: kalle, Lokal_Profil.

@kalle Would you mind looking over the settings especially in terms of CPU so that it doesn't adversely affect the other sites running from the same server.

We could do something with logrotate if we want to.

I did this since even if the log isn't saved running the server, Gradle saves logs elsewhere. This is apparently a known issue:

Lokal_Profil moved this task from Waitin' to Done on the User-Sebastian_Berlin-WMSE board.

@kalle rebuilt the service and re-deployed. Seems to be working fine now

CPU is at nothing when idle now. I upgraded to Java 17 (from Java 11) and replaced the old Google Guice and Jetty framework with a new Spring and Tomcat framework. I still don’t know why this helped, our own business logic was left as is.

It feels stable though.

Also set the log to info level rather than trace/debug, so it doesn’t produce that much output unless there’s an actual error.