Proposal for High Availability configuration + checkpoint keepalived config and monitoring code

[ramr]

Proposal to add HA sidecar configuration via a new admin command:
openshift admin ha-config [<name>] <options>
More details and usage scenarios are in https://github.com/ramr/origin/blob/hasidecar/docs/proposals/ha-configuration.md

The upshot of this proposal is to allow an administrator to specify a HA configuration for a set of nodes (retrieved based on the selector option passed). These nodes would then run a HA sidecar service in a privileged container and generate the keepalived configuration, cluster the keepaliveds and distribute VIPs thru' the cluster. The VIPs themselves will be only enabled when the monitored service is available (the most common use case would be the HAProxy router serving port 80). The HA sidecar container would also monitor the target service and automatically failover the VIPs on failures.

This PR also checkpoints a cleaned up version of the keepalived demo pieces for subsequent use.

@pweil- @smarterclayton can you PTAL.

Thanks.

[smarterclayton]

Cool, will look.

----- Original Message -----

Proposal to add HA sidecar configuration via a new admin command:
openshift admin ha-config [<name>] <options>
More details and usage scenarios are in
https://github.com/ramr/origin/blob/hasidecar/docs/proposals/ha-configuration.md

The upshot of this proposal is to allow an administrator to specify a HA
configuration for a set of nodes (retrieved based on the selector option
passed). These nodes would then run a HA sidecar service in a privileged
container and generate the keepalived configuration, cluster the
keepaliveds and distribute VIPs thru' the cluster. The VIPs themselves will
be only enabled when the monitored service is available (the most common
use case would be the HAProxy router serving port 80). The HA sidecar
container would also monitor the target service and automatically failover
the VIPs on failures.

This PR also checkpoints a cleaned up version of the keepalived demo pieces
for subsequent use.

@pweil- @smarterclayton can you PTAL.

Thanks.
You can view, comment on, or merge this pull request online at:

#1730

-- Commit Summary --

• Add new ha-config keepalived failover service container code, image and
• Add HA configuration proposal.

-- File Changes --

A docs/proposals/ha-configuration.md (186)
M hack/build-images.sh (2)
A images/ha-config/keepalived/.gitignore (2)
A images/ha-config/keepalived/Dockerfile (17)
A images/ha-config/keepalived/README.md (312)
A images/ha-config/keepalived/bin/.gitignore (2)
A images/ha-config/keepalived/conf/hello-openshift-template.json (63)
A images/ha-config/keepalived/conf/settings.sh (78)
A images/ha-config/keepalived/lib/config-generators.sh (270)
A images/ha-config/keepalived/lib/failover-functions.sh (39)
A images/ha-config/keepalived/lib/utils.sh (132)
A images/ha-config/keepalived/makefile.test (11)
A images/ha-config/keepalived/monitor.sh (15)
A images/ha-config/keepalived/tests/echoserver.py (36)
A images/ha-config/keepalived/tests/verify_failover_image.sh (90)

-- Patch Links --

https://github.com/openshift/origin/pull/1730.patch
https://github.com/openshift/origin/pull/1730.diff

---

Reply to this email directly or view it on GitHub:
#1730

[smarterclayton]

Proposal looks good, going to dive into details today.

On Apr 14, 2015, at 1:43 PM, ramr notifications@github.com wrote:

Proposal to add HA sidecar configuration via a new admin command:
openshift admin ha-config []
More details and usage scenarios are in https://github.com/ramr/origin/blob/hasidecar/docs/proposals/ha-configuration.md

The upshot of this proposal is to allow an administrator to specify a HA configuration for a set of nodes (retrieved based on the selector option passed). These nodes would then run a HA sidecar service in a privileged container and generate the keepalived configuration, cluster the keepaliveds and distribute VIPs thru' the cluster. The VIPs themselves will be only enabled when the monitored service is available (the most common use case would be the HAProxy router serving port 80). The HA sidecar container would also monitor the target service and automatically failover the VIPs on failures.

This PR also checkpoints a cleaned up version of the keepalived demo pieces for subsequent use.

@pweil- @smarterclayton can you PTAL.

Thanks.

You can view, comment on, or merge this pull request online at:

#1730

Commit Summary

Add new ha-config keepalived failover service container code, image and
Add HA configuration proposal.
File Changes

A docs/proposals/ha-configuration.md (186)
M hack/build-images.sh (2)
A images/ha-config/keepalived/.gitignore (2)
A images/ha-config/keepalived/Dockerfile (17)
A images/ha-config/keepalived/README.md (312)
A images/ha-config/keepalived/bin/.gitignore (2)
A images/ha-config/keepalived/conf/hello-openshift-template.json (63)
A images/ha-config/keepalived/conf/settings.sh (78)
A images/ha-config/keepalived/lib/config-generators.sh (270)
A images/ha-config/keepalived/lib/failover-functions.sh (39)
A images/ha-config/keepalived/lib/utils.sh (132)
A images/ha-config/keepalived/makefile.test (11)
A images/ha-config/keepalived/monitor.sh (15)
A images/ha-config/keepalived/tests/echoserver.py (36)
A images/ha-config/keepalived/tests/verify_failover_image.sh (90)
Patch Links:

https://github.com/openshift/origin/pull/1730.patch
https://github.com/openshift/origin/pull/1730.diff
—
Reply to this email directly or view it on GitHub.

[ramr]

bump - @smarterclayton any comments/updates. There's an implementation PR #1807 behind this one. Thx

[smarterclayton]

I added quite a lot of comments to #1807

[ramr]

Aargh, my bad. Didn't see those as yet - will address those on #1807 . Thanks.

[ramr]

@smarterclayton if you can please merge this, then I can rebase it in PR #1807 , make the necessary modifications in that subsequent PR (and address your comments therein) and squash the commits. Thanks.

[smarterclayton]

Why do you need that? You can just close this, since these commits are in your other pull

[ramr]

Ok - did have to do a 2 step rebase to completely get this in . Closing this out.