search

metal-stack / metal-hammer

metal-hammer is used to boot bare metal servers with ipxe and the metal-stack kernel
GNU Affero General Public License v3.0
41 stars 6 forks source link

only reboot every 3 days and spread in one day #75

Closed majst01 closed 2 years ago

majst01 commented 2 years ago

If a lot of machines powered on at the same time, they will also reboot every 24 hours at the exact same time which causes a lot of link up / link down events on the switches. This might as a consequence stress the switchd on the switch itself and eventually cause some small network hiccups for services connected to this switch.

From test environment:

INFO[04-14|05:16:31] autoreboot set to                        after=72h0m0s spread=24h0m0s caller=kernel.go:141                                                                                                                                                                           
INFO[04-14|05:16:31] autoreboot with spread                   after=95h57m0s caller=kernel.go:150
Gerrit91 commented 2 years ago

Three days appears pretty long to me. When you roll out a new feature through metal-hammer it takes three days until it's becoming active for the end user without the provider having a nice opportunity to restart all waiting machines for large partitions.