Suggestions about gluster-kubernetes HOT 6 CLOSED

@jarrpa @nick yeah /var/lib/heketi will be present in all the gluster pod running nodes. That's where we persist the mount points for the bricks created for each volume. yeah abort can clean it up, will not cause any problem. ONLY when this hits we are safe to remove it.

from gluster-kubernetes.

Thank you for your participation! :) To address your suggestions:

• Good idea. I need to check the output of the heketi-cli commands for errors as well, so I'll do both those things at around the same time.
• I'm not sure what you mean by this one. /var/lib/heketi shouldn't exist on the kube nodes, the GlusterFS volume heketi uses for its database should be mounted to /var/lib/heketi within the heketi container. At best I could delete the contents of the volume, but that would still leave the volume around, which would be a problem.
• I don't think I want to be messing directly with extant VGs in the deployment script. I'd have to be very careful to make sure I'm not deleting anything I shouldn't be, and likely prompting confirmation, at which point it's not much different from telling the admin to delete the VGs themselves.
• Throwing in systemctl calls makes the script less agnostic to linux distro, which is something I'm aiming for (even if I'm currently only developing in RH-flavored distros) :)

I personally use the vagrant setup for my development and testing. There's a wonderful little plugin called 'sandbox' (vagrant plugin install sandbox) that manages a simple snapshot-based workflow:

./up.sh -> vagrant sandbox on -> <do stuff, mess it up> -> vagrant sandbox rollback -> <restart docker service on all the nodes, or just restart the VMs> -> <do stuff again>

which makes things fairly easy to automate.

from gluster-kubernetes.

It seems that /var/lib/heketi on each node may be a leftover from the bootstrap heketi installation perhaps? Regardless, if I run --abort then I'll still have to manually rm -rf /var/lib/heketi on each node before running the script again.

from gluster-kubernetes.

I went looking and I see them in my setups too, but I've never had to rm the directories during my development. I can look into whether or not this is a good idea, depending on when the directories get made. After what point in the script's execution are you running the --abort?

from gluster-kubernetes.

@MohamedAshiqrh That makes sense! :)

from gluster-kubernetes.

As all the original suggestions have been addressed or discussed, closing this issue.

from gluster-kubernetes.