Power down before outages
Jump to navigation
Jump to search
Name of Task: Power down before outages
People Involved: Craig Earley, Davit Kvartskhava
Date Started: June 5, 2018 (or thereabouts)
Deadline:
Description:
After a hard crash in June 2018, we want to have a program that powers down our servers gracefully after we switch to battery power. The program APCUPSD should be helpful.
Phase 1
APCUPSD is installed and works. It does not yet interface with our hardware.
We have useful SNMP codes as well, which can be helpful in getting information.
Some notes scraped from a Google Drive Doc:
- Decided we’ll use the current UPS info and adapt later if needed
- Send message to ask machines to initiate shutdown
- Apcupsd: need some more info to actually connect to the UPS, but the software runs on shinken
Phase 2 - current status
We've decided to take a different approach to fulfill this project.
- Decided not to use APCUPSD even though it is a great software for controlling the UPS. The wide range of functionality that it offers is simply not needed in our case. We just need to get the info about the current battery level.
- The current battery level can be extracted using SNMP.
- Another important aspect: In the sake of good network design and particularly, robustness, we determined that it is better not to have a centralized system that asks other machines to initiate the shutdown. Instead, each machine will make a decision on its own, based on the level of the battery of the UPS which will be the same for all.
- Each machine will run a python file periodically using the Linux daemon Cron[1].
- Things that need to be discussed:
- What should be the time interval for checking the battery power?
- Whats the critical battery level?
- How are we going to test on a single machine at first?
- Where should I keep the python file?
Helpful sources
Date Last Updated:' August 1, 2018
