I did not realize there was a management operation in the mix. I agree with Robbie
@Havret let me know if that’s the issue please ? On Tue, Oct 26, 2021 at 6:05 AM Robbie Gemmell <robbie.gemm...@gmail.com> wrote: > If the management API has a queue deletion operation that takes an > 'auto delete address' boolean then I would consider it a bug if it > doesnt actually [eventually] delete a previously > auto-created-by-management address when that boolean is true, > regardless of the broker config. > > If it doesnt delete the address then it seems effectively useless; if > it only does anything if auto-delete-address was set true in the > broker config then why does the management boolean even exist at all? > It then seems to do nearly or absolutely nothing (unless setting the > management boolean false will somehow prevent the address being > auto-deleted by the broker even if its auto-delete config is set > true...seems unlikely)? > > Personally I think this is a bug and the explicit management boolean > should be respected, it and the broker auto deletion are not even > really the same form of auto-deletion. The general auto-deletion > mechanism is a sweep recognising unused auto-created things. The > management operation is an explicit command that 'im deleting the > queue, also auto delete its address now, the one that you auto-created > earlier, so that I dont have to go do that separately'. It is > essentially a shortcut to avoid needing to perform another management > operation. > > At the very least if it is to be considered not-a-bug it should throw > an exception when the value is true if it isnt actually going to work, > e.g an IllegalStateException, alerting the user to their folly of > expecting it to work in the obvious way it previously seems to have. > > In reverse, I would also expect/assume an 'auto create address' > management boolean on queue creation to work, even when the 'auto > create addresses' broker config is set false. If not, again it should > throw. > > On Tue, 26 Oct 2021 at 06:39, Havret <h4v...@gmail.com> wrote: > > > > It seems I was assuming that calling destroy queue via management API has > > greater precedence than configuration setting. With that being said I > will > > just update the client docs, so the API doesn't confuse users anymore. > > > > To summarize: > > If you create a queue with *auto-create-address *flag set to true, and > then > > try to destroy it, the flag *autoDeleteAddress *will work only, and only > if > > the broker configuration has *auto-delete-addresses* flag set to true. > > > > On Tue, Oct 26, 2021 at 12:31 AM Clebert Suconic < > clebert.suco...@gmail.com> > > wrote: > > > > > If auto-delete is disabled, the address should not be removed. > > > > > > If you had a case where it was removed, it was a bug. > > > > > > The flag was respected before.. but perhaps there was a situation it > was > > > not. > > > > > > > > > You telling me to disrespect the flag and always remove it.. would be > > > an oxymoron.. I would just pretty much ignore the setting. > > > > > > On Mon, Oct 25, 2021 at 4:56 PM Havret <h4v...@gmail.com> wrote: > > > > > > > > This particular queue is not defined in broker.xml. I'm not sure if > 2.19 > > > is > > > > semantically correct, as having the same configuration I'm getting > > > > different results than with 2.18 and previous versions. I understand > that > > > > removing the address was decoupled from removing the queue, but in > this > > > > particular case the address is never deleted. > > > > > > > > I'm gessing that changes related to > > > > https://issues.apache.org/jira/browse/ARTEMIS-3502, > > > auto-delete-addresses > > > > flag, and management API delete queue (with autoDeleteAddress set to > > > true) > > > > don't play nice together. > > > > > > > > My expectation would be that the address will be eventually deleted > even > > > > though the auto-delete-addresses flag is disabled in broker > > > configuration. > > > > This is an imperative API call, at least from the client perspective. > > > > > > > > On Mon, Oct 25, 2021 at 5:46 AM Clebert Suconic < > > > clebert.suco...@gmail.com> > > > > wrote: > > > > > > > > > I believe 2.19 is semantically correct to the configuration you are > > > > > showing. > > > > > > > > > > We also had another change related to not delete queues defined on > > > > > broker.xml. Perhaps that’s the problem ? > > > > > > > > > > > > > > > > > > > > On Sun, Oct 24, 2021 at 9:38 AM Havret <h4v...@gmail.com> wrote: > > > > > > > > > > > Hi Clebert, > > > > > > > > > > > > I had the same broker configuration for 2.18.0 and the test > passed > > > just > > > > > > fine, so I don't think that the matter of new defaults. The > queue is > > > > > > removed just fine ( > > > > > > > > > > > > > > > > > > > > > https://github.com/Havret/dotnet-activemq-artemis-client/blob/6cee91547c18194330062a48c3b02c7c385457ff/test/ArtemisNetClient.IntegrationTests/TopologyManagement/DeleteQueueSpec.cs#L68 > > > > > > ), > > > > > > the problem is that the management api seems to ignore > > > autoDeleteAddress > > > > > > parameter and the second assertion fails ( > > > > > > > > > > > > > > > > > > > > > https://github.com/Havret/dotnet-activemq-artemis-client/blob/6cee91547c18194330062a48c3b02c7c385457ff/test/ArtemisNetClient.IntegrationTests/TopologyManagement/DeleteQueueSpec.cs#L70 > > > > > > ) > > > > > > > > > > > > Thanks, > > > > > > Havret > > > > > > > > > > > > > > > > > > On Thu, Oct 21, 2021 at 2:54 AM Clebert Suconic < > > > > > clebert.suco...@gmail.com > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > The answer for that is in your configuration: > > > > > > > > > > > > > > <auto-delete-queues>false</auto-delete-queues> > > > > > > > <auto-delete-addresses>false</auto-delete-addresses> > > > > > > > <auto-delete-created-queues>false</auto-delete-created-queues> > > > > > > > > > > > > > > > > > > > > > if you create a new broker configuration now, these > auto-deletes > > > are > > > > > > > turned off. > > > > > > > > > > > > > > > > > > > > > I did that because I have been recommending people in > production to > > > > > > > turn them off.. so I thought I should just turn'em off by > > > default... > > > > > > > > > > > > > > I couldn't change the default default, but I added these to new > > > > > > > configurations... > > > > > > > > > > > > > > > > > > > > > if you need to test auto-delete, you need to set them true. > > > > > > > > > > > > > > > > > > > > > > > > > > > > Notice that since 2.19 now these will perform a scan before an > > > actual > > > > > > > delete would happen. > > > > > > > > > > > > > > This is because I had seen a production issue where a consumer > that > > > > > > > was up to date, disconnected (connection failure), the close > > > deleted > > > > > > > the queue immediately, while immediately it caused to another > > > queue to > > > > > > > be created... in one cirtcunstance the queue was created but > not > > > the > > > > > > > address, creating a mess.. > > > > > > > > > > > > > > > > > > > > > I tried to fixing it properly without adding the delays, but I > > > could > > > > > > > not come up with a scenario that would always work perfectly.. > > > because > > > > > > > the client will query for the queue, and then issue a create > > > > > > > separately.. there's always an issue between the queue and > > > create... > > > > > > > so... the best is to not remove the queue immediately any > more... > > > > > > > > > > > > > > On Wed, Oct 20, 2021 at 4:33 PM Havret <hav...@apache.org> > wrote: > > > > > > > > > > > > > > > > Hi, > > > > > > > > > > > > > > > > I'm trying to update the test suite of ArtemisNetClient to > use > > > > > ActiveMQ > > > > > > > > Artemis 2.19.0, but one of the tests starts failing on > 2.19.0: > > > > > > > > Should_delete_queue_alongside_with_auto_created_address --> > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > https://github.com/Havret/dotnet-activemq-artemis-client/pull/300/checks?check_run_id=3943848841 > > > > > > > > > > > > > > > > The test creates a queue using the management api and then > > > removes it > > > > > > > with > > > > > > > > flag autoDeleteAddress set to true. The queue is removed but > the > > > > > > address > > > > > > > > isn't. I've looked over the release notes, but I didn't spot > > > anything > > > > > > > that > > > > > > > > could indicate this change of behavior. Is this a > regression, or > > > was > > > > > > it a > > > > > > > > deliberate change? > > > > > > > > > > > > > > > > Robbie suggested that it might be related to > > > > > > > > https://issues.apache.org/jira/browse/ARTEMIS-3502 > > > > > > > > > > > > > > > > I tried to adjust the test: > > > > > > > > - I set *address-queue-scan-period* to 100 > > > > > > > > - I wait up to 5 seconds until the address is removed > > > > > > > > > > > > > > > > But the test still is failing. My broker configuration is as > > > follows: > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > https://github.com/Havret/dotnet-activemq-artemis-client/blob/fc256b2a3784447fa4feadea58b62f58d1fcabfd/test/artemis/broker.xml > > > > > > > > Is it possible that this new setting interferes with > > > > > > > > *auto-delete-addresses *set to *false*? > > > > > > > > > > > > > > > > Thanks, > > > > > > > > Havret > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > > Clebert Suconic > > > > > > > > > > > > > > > > > > -- > > > > > Clebert Suconic > > > > > > > > > > > > > > > > > -- > > > Clebert Suconic > > > > -- Clebert Suconic
