Retiring Webportal Servers
Two reasons one might need to retire a webportal server:
- 1.Scaling limitations
- 2.Takedown requests
One of the core processes that makes Skynet possible, and is a key component of the webportal stack, is the
skyddaemon. There are a few known limitations of the
skyddaemon that have been observed that can impact the performance of the webportal. These limitations are:
- 1.Contract Size > 100TB
- 2.Repair Data > 10 TB
- 3.Total Number of Files > 1.5 million
If one of these conditions is met, it is recommended to retire the server and put it in maintenance mode. This means simply removing it from your Load Balancer so that it is not serving new traffic, but is still running and maintaining the data it has pinned.
To do this, you simply must add the server to the
out_of_LBgroup in your
hosts.inifile. This will allow you to still include the server in upgrades, but it will not enable the
health-checkwhich means it will not be added back into the Load Balancer.
If the server is ready to come out of maintenance mode, because the scalability was improved or the repair data is back under control, just remove it from the
out_of_LBgroup and run the deploy or restart script.
Some takedown requests are not immediately resolvable. In order to comply with a hosting provider and avoid them turning off access to the server, you can use the takedown-skynet-webportal.sh script. This will shut down all services except
skyd, which will prevent external traffic but allow the
skyddaemon to keep the data healthy on the network.
When the issue is resolved and the server is ready to be brought back up, simply remove the server from the
webportals_takedowngroup and run the deploy or restart script.