I wrote a go script that connects to the Juju Log endpoint, and sends messages (as fast as possible) for 5s. (see attached)
With that, I was able to spawn 100 machine agents, all running concurrently.
Note that the logsender explictly doesn't call Writer.Close(), though I haven't done extensive testing if that matters.
After running 100 at a time multiple times, I do see memory growth. I can't quite tell if it is only peak-memory based on max number of concurrent loggers. Still investigating.
I wrote a go script that connects to the Juju Log endpoint, and sends messages (as fast as possible) for 5s. (see attached)
With that, I was able to spawn 100 machine agents, all running concurrently.
Note that the logsender explictly doesn't call Writer.Close(), though I haven't done extensive testing if that matters.
After running 100 at a time multiple times, I do see memory growth. I can't quite tell if it is only peak-memory based on max number of concurrent loggers. Still investigating.