Task #2082
closed
provide a way to set per-vm backup time (e.g. only do gerrit during nighttime)
Added by Christian Lohmaier over 7 years ago.
Updated over 3 years ago.
Target version:
Team - Q2/2020
Description
Right now, salt does random minute/hour assignments, but those are suboptimal when it comes to backing up the produciton machines, as it can happen that the backup happens while the machine is already doing much i/o, so rsync scanning everything is a big, unnecessary penalty...
salt should read corresponding pillar data and use that and only use auto-date when there is none.
Has there been any progress? IMHO it makes sense to involve Guilhem at this stage, so he gets insight
the salt template tdf/rsnapshot/crontab just contains:
{% for host, hostinfo in salt['mine.get']('*', 'grains.items').items() %}
{% set minute = salt['random.seed'](60, hostinfo['id']) -%}
{% set hour = salt['random.seed'](24, hostinfo['id']) -%}
{{ minute }} {{ hour }} 2-6,8-13,15-20,22-27,29-31 * * root /usr/bin/rsnapshot -c/etc/rsnapshot/{{ hostinfo['id'] }}.conf sync; /usr/bin/rsnapshot -c/etc/rsnapshot/{{ hostinfo['id'] }}.conf daily
[similar lines for weekly/monthly]
{% endfor %}
in other words for all hosts it pics a random minute an hour to run the backup. For machines that take a hit when being backed up, it is desireable to have a pillar data used instead. so a pillar value "backup-time: minute hour" should be used if set, and only use tha random minute/hour if the pillar data is not set.
- Assignee changed from Christian Lohmaier to Guilhem Moulin
Reassigning to Guilhem, so he can have a look
This ticket is 1,5 years old - is it relevant anymore at all?
Yes, keeping that open as I planing to heavily work on the backup solution in Q2/Q3 (cf. minutes from the Jan 16 infra call).
- Target version changed from Pool to Q4/2019
Any update? Can this be closed?
- Target version changed from Q4/2019 to Q2/2020
- Status changed from New to Closed
Also available in: Atom
PDF