Conversation
|
|
||
| @Override | ||
| public void start() { | ||
| public void invoke() { |
There was a problem hiding this comment.
Changed this (internally used) method name, as it's always a blocking operation and invoke seems more accurate.
There was a problem hiding this comment.
that's a good change, I was confused by getDockerCompose(...).start(...) constructions :)
P.S. do we use anything else besides invoke() on getDockerCompose's return value? Is not, maybe we can simplify it to something like runWithCompose("down -v") ?
Reduce unnecessary cleanup attempts which cause errors to be logged. docker-compose down is now trusted to have cleanup up properly if it exits with a 0 status code.
3944eac to
a7dbb82
Compare
| .getExitCode(); | ||
|
|
||
| if (exitCode == null || exitCode != 0) { | ||
| throw new ContainerLaunchException( |
There was a problem hiding this comment.
As with LocalDockerCompose, we now throw an exception if a failure occurred. This allows us to detect a failure of docker-compose down.
| try { | ||
| try { | ||
| // First try to remove by name | ||
| dockerClient.removeNetworkCmd(networkName).exec(); |
There was a problem hiding this comment.
This seems to always fail and causes an error log. I'm just running some checks that this doesn't let any excess containers survive cleanup.
There was a problem hiding this comment.
@rnorth ouch. It doesn't fail when recently added Docker Networks support is being used (because we know the IDs), so I wouldn't remove it.
Since we list networks in Docker Compose's implementation, maybe we can use the IDs there as well?
| // If we reach here then docker-compose down has cleared networks and containers; | ||
| // we can unregister from ResourceReaper | ||
| spawnedNetworkIds.forEach(id -> ResourceReaper.instance().unregisterNetwork(identifier)); | ||
| spawnedContainerIds.forEach(id -> ResourceReaper.instance().unregisterContainer(id)); |
There was a problem hiding this comment.
This is the core of the change - we unregister containers and networks if docker-compose down seems to have worked.
|
This should fix #342 |
| try { | ||
| try { | ||
| // First try to remove by name | ||
| dockerClient.removeNetworkCmd(networkName).exec(); |
There was a problem hiding this comment.
@rnorth ouch. It doesn't fail when recently added Docker Networks support is being used (because we know the IDs), so I wouldn't remove it.
Since we list networks in Docker Compose's implementation, maybe we can use the IDs there as well?
| // we can unregister from ResourceReaper | ||
| spawnedContainerIds.forEach(id -> ResourceReaper.instance().unregisterContainer(id)); | ||
| spawnedNetworkIds.forEach(id -> ResourceReaper.instance().unregisterNetwork(identifier)); | ||
| } catch (ContainerLaunchException e) { |
There was a problem hiding this comment.
maybe catching just Exception will be better here? Any Exception in invoke will prevent Reaper from being called
| ResourceReaper.instance().removeNetworks(identifier); | ||
| // If we reach here then docker-compose down has cleared networks and containers; | ||
| // we can unregister from ResourceReaper | ||
| spawnedContainerIds.forEach(id -> ResourceReaper.instance().unregisterContainer(id)); |
There was a problem hiding this comment.
FYI spawnedContainerIds.forEach(ResourceReaper.instance()::unregisterContainer); is also valid here :)
There was a problem hiding this comment.
Not quite sure I follow this:
Since we list networks in Docker Compose's implementation, maybe we can use the IDs there as well?
But I'd interpret this as 'let's only register network IDs, not names'. Is that right? I'll change and make sure we use IDs throughout, then I think we should be OK.
|
|
||
| @Override | ||
| public void start() { | ||
| public void invoke() { |
There was a problem hiding this comment.
that's a good change, I was confused by getDockerCompose(...).start(...) constructions :)
P.S. do we use anything else besides invoke() on getDockerCompose's return value? Is not, maybe we can simplify it to something like runWithCompose("down -v") ?
|
|
||
| // Compose can define their own networks as well; ensure these are cleaned up | ||
| dockerClient.listNetworksCmd().exec().forEach(network -> { | ||
| if (network.getName().contains(identifier)) { |
There was a problem hiding this comment.
not related to this PR, but AFAIK we shouldn't use contains here, but startsWith or whatever the rule for Docker Compose, otherwise if identifier if "a", then it will delete all networks with "a" character in name :D
| * @param networkName the image name of the network | ||
| * @param id the ID of the network | ||
| */ | ||
| public void registerNetworkForCleanup(String networkName) { |
There was a problem hiding this comment.
Since this method was semi-public, maybe it worth keep the old one and mark as deprecated, at least until 1.5.0?
There was a problem hiding this comment.
Yep - will reinstate all public methods as they were and deprecate, 👍
| private void removeNetwork(String networkName) { | ||
| private void removeNetwork(String id) { | ||
| try { | ||
| try { |
There was a problem hiding this comment.
I think this block should do the job when parameter is id, and "list with filter" is not required anymore
There was a problem hiding this comment.
It could be clearer - I'll do that.
I think we still need to keep the list-with-filter operation though: if the network has already been removed for any reason, dockerClient.removeNetworkCmd will log out an error. So we need to do something to check for the existence of the network first that won't log an error.
Improve comments to aid clarity in removeNetwork method
Reduce unnecessary cleanup attempts which cause errors to be logged.
docker-compose down is now trusted to have cleanup up properly if it
exits with a 0 status code.
I hope we can squeeze this in to 1.4.0!