Git shared repository: Difference between revisions
No edit summary |
No edit summary |
||
Line 70: | Line 70: | ||
* 2016-09-07 | * 2016-09-07 | ||
* 2018-04-16 | * 2018-04-16 | ||
(I did it but did not get logged here) | |||
* 2020-05-27 |
Latest revision as of 01:42, 28 May 2020
The git shared repositories used by the genome browser staff and genecats
is located at hgwdev:/data/git
Inside there are various bare repositories including kent.git. See this Getting Started With Git page for more info.
There are cron jobs that update various copies of kent.git such as the hourly copies that are refreshed with git pull, and a copy which is updated every 15 minutes that the general public fetches from, both directly with git, and with web-git. In addition the admins keep daily, weekly, monthly, and yearly backups of hgwdev:/data/git.
From Galt's crontab:
# update git public repo 0,15,30,45 * * * * /data/git/public/refresh.csh 0 * * * * /data/git/backup/refresh.csh
Periodically (at least every 6 months, but maybe more often)
git needs to garbage-collect the shared bare repository.
Not ONLY to save space, it ends up saving time too.
It had grown to 2GB! After gc, it was only 288MB.
Also, it had taken git bare clone took 13 minutes to run.
After gc, the git bare clone took less than 1 second.
And then git regular clone was only 3 seconds!
# Send out announcement to genecats and browser-staff # that we will be doing git maintenance for 5 minutes. ssh hgwdev cd /data/git # move the real repo aside so nobody can use it. mv kent.git kent-real.git # get rid of the old copy if any rm -fr kent.git.copy # make a new backup copy just to be safe rsync -a kent-real.git kent.git.copy cd kent-real.git # only takes 5 minutes git --bare gc cd .. mv kent-real.git kent.git
That's it! It doesn't seem that the other git repos in /data/git/ get enough commits to even bother with gc.
Apparently, when we run a git on a non-bare clone such as your personal ~/kent, it does gc automatically.
It is preferable to perform this garbage collection late at night when few people are working.
Garbage Collection Done:
- 2012-07-23
- 2013-03-26
- 2013-07-29
- 2014-08-08
- 2015-04-03
- 2016-01-04
- 2016-09-07
- 2018-04-16
(I did it but did not get logged here)
- 2020-05-27