pulsar

And I have verified by MQTTBox, it works.

Hi all, I created an issue <https://github.com/streamnative/mop/issues/1034>. Can anyone help answer if MOP does not currently support shared subscription?

if you push a fix，please keep all the tests passed

The link above is a bug report, the PR itself has been closed. We ran into a number of issues regarding that PR

I can outline the issues in the strategy we tried

for this issue，could you share the reproduce codes？

That specific issue is pretty much reproducible any time we setup clustered pulsar with MoP Proxy

<https://github.com/streamnative/mop/pull/957>

i don’t believe you have tested in your env

so for issues, please detail the steps or reproduce steps first

MQTT keep alive only sends PINGREQ in absence of other control messages. The keepalive mechanism in the proxy forwards PINGREQ as a way to keep adapterChannels alive, so if you are interacting on one topic you are interacting with only one broker, and the client might not even send PINGREQ meaning the other adapterChannels are closed by the brokers IdleStateHandlers

Ticket #1212 has both steps to reproduce, as well as a description of how and why the bug occurs

if the test can’t pass, then we can discuss the issue

the mop have tests and can support multi-brokers or only brokers.

you should add a test for your description

In previous PRs I've asked for input on exactly how you'd want these tests written, or where exactly these tests should go. You do not have contribution guidelines, and your tests do not include many comments, and the tests are split between 2 different directory structures. You had never responded.

i think  you can add test in the tests dir

you can push the fix ,so  i think adding test is more simple

or any package , but it can help reproduce the issue

I have steps to reproduce the MoP Proxy issue

The PR for that ticket I've abandoned. I don't have intentions to fix this, because I'm either not understanding how the proxy is supposed to work or it simply does not actually work

Am I understanding that you want integration tests to reproduce the issue?

even we push the fix, we still need the tests to verify this.

Do you understand the description I've provided for the issue?

The particular area for concern is here: <https://github.com/streamnative/mop/blob/master/mqtt-impl/src/main/java/io/streamnative/pulsar/handlers/mqtt/proxy/MQTTProxyProtocolMethodProcessor.java#L222|https://github.com/streamnative/mop/blob/master/mqtt-impl/src/main/java/io/streamn[…]ulsar/handlers/mqtt/proxy/MQTTProxyProtocolMethodProcessor.java>

normally, the mop will create one connection for every client, this will take a lot of resources. so we design a new protocol internally to support one connection can support more clients.

`topicBrokers.values().forEach(adapterChannel -&gt; {`
            `adapterChannel.thenAccept(channel -&gt; {`
                `msg.setClientId(clientId);`
                `channel.writeAndFlush(msg);`
            `});`
        `});`

you forward a pingreq per topic that has an adapterChannel associated with it

If I create a client and create 8 subscriptions, each PINGREQ gets 8 PINGRESP

Adapter Channels aren't shared between proxy connections

proxy don’t need send more ping to the broker, it will create one connection to the broker, so even if there are 8 clients, the clients will send 8 pings and receive 8 pongs between the proxy, but the proxy will send one ping to the broker

I spent 3 weeks observing packets in wireshark and watching the process in a debugger

I can try to write integration tests reproducing the issue, but it won't be easy, and I can't assure you that I will have time. I mean I have a job also, and I have a lot of things to do as I'm sure you do as well. At the end of the day, I can only dedicate as much time as my employer is willing to allocate

To put it bluntly, it's seems quite odd to me that a representative of a company talking to a prospective customer would make such an ask. I've made 4 contributions already to MoP. I have maintained several OSS projects. I'm generally much more responsive, and frankly much more kind to the people who take their own time to help me on those projects. I'd personally approach situations in talking to potential customers or stakeholders much more professionally.

yes, i can try to write the test later.
&gt; I’ve made 4 contributions already to MoP
really thanks for your contribution, but you should add test to insure the logic and stable .  and keep all the current test passed.   <https://github.com/streamnative/mop/pull/957>  has break the logic, I have fixed that in the later patches. so for this issue,  i suggest you to add the reproduce codes.

and  for the proxy logic you point out, it’s a improvement we have made.  you’d better to try test these tests or  take a look at the issue or motivations we describe in the patch.

For this particular issue you had asked for tests, I had asked for clarification, and someone else merged the PR without following up. I'm fine adding or adapting tests, and I understand the importance. The issue is that you responded asking for tests 2 months later, and then I asked for clarifications and hadn't heard anything back.

Usually when this happens on projects I maintain I'll point to the test that failed and maybe give insight to what failed. If I believe new tests need to get added I'll usually pass along a rough expectation of what that test does or should do.

Also this issue is completely unrelated to the current issue, and you keep bringing it up. I broke the tests in this one case, I wasn't the one who merged and accepted it. I was fully willing to comply with any reasonable change requests provided sufficient details. I've made 4 contributions to fix pretty substantial bugs. LWT and retain messages wouldn't even reach Pulsar topics previously. I'm not going to sit here and point out these flaws, we've moved past that.

Frankly, I don't have any more time permitted to contribute to a PR to fix the issues I've seen in the MoP Proxy. You could be helpful and maybe asked direct detailed questions about what exactly you want in terms of details to reproduce the issue. I've gone through the source code and already detailed what I believe is happening, which is far beyond what most bug reports would do. However, you don't seem satisfied with that, and instead of asking for clarifications or details, you are making a large request that I fix your tests to catch an issue you have failed to catch originally. Mistakes happen, bugs are part of the software lifecycle, but I think you ask is simply unreasonable. You could be doing a lot more to be involved in this, and instead you are bringing up a past issue that is completely irrelevant, and suggest that I make a large effort to provide integration tests to prove the issue.

It is not my job, nor my intention to continue to be a major contributor to MoP. The experience in doing so has been outstandingly bad, had the relationship been better we'd likely have been happy to contribute. Our evaluation has led us to look towards other solutions, which is a shame, because MoP standalone has been working fine for us after we had patched several of the issues we originally found, but we are currently limited to running standalone pulsar, because of these issues.

this is a connection with 2 subscriptions, as per the code a PINGRESP is received per topic that's been interacted with on the connection, this is running this image <https://hub.docker.com/layers/streamnative/sn-pulsar/3.1.2.1/images/sha256-a9b8363247ecd24c554640d1b8ef04f467f2d120a5b9daf40a9531fd0556c668?context=explore|https://hub.docker.com/layers/streamnative/sn-pulsar/3.1.2.1/images/sha256-a9b836324[…]640d1b8ef04f467f2d120a5b9daf40a9531fd0556c668?context=explore>

Here are full steps to reproduce including a minikube setup with an example Java program. I recommend you observe everything in wireshark to see why it fails. <https://github.com/TTCColorado/mop-proxy-bug-reproduction>

Follow the readme, this should setup Pulsar using the apache pulsar helm chart, and runs the sn-pulsar container image which has the 3.1.2.1 MoP release

Please let me know if you have trouble running that.

Hi <@U06EC448ZK3>
<https://github.com/streamnative/mop/pull/1287>
I have left some comments for your fix.
Thanks for find this issue.

There is a closed PR for the fixes I attempted