notifications are now system level instead of user level.
To clarify the use events/notifications/email:
* eventlog - everything that is happenning on server
* notifications - specific important events (alerts)
* email - these are really urgent things that require immediate attention. this is for
the case where an admin does not visit the dashboard often. can also be alerts like
bad backup config or reboot required which are not events per-se.
Notes on notifications
* oom - notification only
* appUpdated - notification only
* cert renewal failure - only raise when < 10 days to go. also send email thereafter (todo).
* Backup failure - only if last 5 backups failed (todo).
* Box update - notification only. we anyway send newsletter.
* box update available - we raise a notification. no email.
* app update available - we already have update indicator on dashboard. so, no notification or email.
Alerts:
* backup config
* disk space
* mail status
* reboot
* box updated
* ubuntu update required
the platform.start can take forever. this means that we start the
clock to include platform.start and this sends a lot of spurious
up/down notifications.
also, bump the down threshold to 20 mins.
We removed httpPort with the assumption that docker allocated IPs
and kept them as long as the container is around. This turned out
to be not true because the IP changes on even container restart.
So we now allocate IPs statically. The iprange makes sure we don't
overlap with addons and other CI app or JupyterHub apps.
https://github.com/moby/moby/issues/6743https://github.com/moby/moby/pull/19001
we can just use container IP instead of all this httpPort exporting magic.
this is also required for exposing httpPaths feature (we have to otherwise
have multiple httpPorts).
because it was a local variable, we were just sending out oom mails
like crazy
also, fixes an issue that if docker.getEvents gets stuck because
docker does not respond then we do not do any health monitoring.
i guess this can happen if the docker API gets stuck.
