mohammedzee1000
commented
6 years ago Operations in action:
LIMIT on images :
openshift_node_kubelet_args={'max-pods': ['150'], 'resolv-conf': ['/etc/resolv.conf'], 'image-gc-high-threshold': ['65'], 'image-gc-low-threshold': ['60']}
Basically, if disk usage crosses 65% GC will come into affect. Note only storage affecting docker is taken into account so if volume is mouted on /var then
only that volumes size is taken into account.
Before cleanup is triggered:
$ docker images -f dangling=true
REPOSITORY TAG IMAGE ID CREATED SIZE
<none> <none> 4967a565b013 3 minutes ago 200 MB
<none> <none> b7e028b3d46f 11 minutes ago 200 MB
registry.centos.org/pipeline-images/ccp-openshift-slave <none> cebfdbb54c6d 26 hours ago 1.28 GB
$ df -h
...
/dev/vdb 59G 28G 29G 50% /var
...
Filling up the space to trigger gc:
Multiple instances of :
$ dd if=/dev/zero of=filex.txt count=104857600 bs=1024
space fill:
$ df -h
/dev/vdb 59G 39G 18G 70% /var
After space fill:
$ docker images -f dangling=true
REPOSITORY TAG IMAGE ID CREATED SIZE
registry.centos.org/pipeline-images/ccp-openshift-slave <none> cebfdbb54c6d 26 hours ago 1.28 GB
Note some images, such as this may not be reoved immediately, if the hash is being used somehere, ebven if its dangling
bamachrn
dharmit
This PR adds docs to enable gc on cluster based on kubelet parameters.
Note: This branch was also placeholder for testing out various approaches, and was hard reset when appropriate approach was selected.
Signed-off-by: Mohammed Zeeshan Ahmed mohammed.zee1000@gmail.com