How does polling work?

Hey Patrick - I just set up a private instance, and am impressed at how easy it was. Thanks for all the thorough documentation!

I’m interested in learning how often the agents poll the web server. And if they’re doing it fairly often, what’s to prevent multiple agents of the same location from retrieving the same job?

Also, it seems like it takes a bit of time (up to 30 minutes) for the list of locations/browsers on the index page to update when I startup or shutdown agents. Is this list cached?

Thanks so much!

BTW - out of curiosity, what makes up the test id? For example: 141126_HK_13CS - I know the first 6 chars are the date. Does the rest of it mean anything?

When an agent is checking for work the server “locks” the work queue (with a file lock in tmp/) to make sure only one thing at a time is operating on the queues.

Starting up agents should show right away (well, as soon as they actually start). If you are spinning up agents on EC2 it can take a good 10 minutes for regular instances and even longer for spot instances.

Shutting down the server doesn’t delete test agents until it has been an hour since they connected which is why it takes a while for them to disappear.

As far as the test ID, the middle is just a random hash (defaults to 2 digits) to prevent the filesystem from throwing up with too many directories. The last block is usually an incrementing counter converted to alphanumeric (or a md5 hash of a random ID for private tests).