When your workers are loading, the normal operation timeout is in
effect. If you spawn many workers or have them perform a lot of work
or wait for just long enough before actually booting, they will be
killed by the worker ping timeout.
If you want to control the exact timeout for your workers to boot,
specify the `worker_boot_timeout` option. Otherwise, it defaults to
the value of the `worker_timeout` option.
Had a customer who thought setting this was preventing against long running requests (it doesn't). These docs clarify the purpose of setting this value, that it is used to detect hung processes.
This allows you to use `tag ""` in the config file to avoid
inferring a tag.
This is useful for example in cluster mode with phased restarts
where you might want the master process to just specify the puma
version and nothing else.
Later in `on_worker_boot` you would manually modify the process
title to suit your needs if you need more specific tags.
The existing umask examples default to no permissions at all, which
doesn't seem particularly useful to anybody. This changes the examples
to read/write, which should be a better baseline. [ci skip]
