If messages published failed with qos=0/1/2, what will mosquitto do?

[ethsonliu]

I dit a test.

A client publish one message with length = 1024 bytes and qos = 1 every 1 second, meanwhile another client subscribe the same topic with the same qos level, and then write to mysql server.

step1: pull up the cable of publish-client for 30 minutes.
step2: reconnect the cable of the publish-client.
step3: wait 5 minutes to let publish-client to process(publish and republish the failed messages)
step4: check the saved datas of mysql server by the timestamp in each message's payload.

The conclusion is: mosquitto will save these published-failed-messages while network is broken, after the network reworks, mosquitto will republish these published-failed-messages, but maybe not all(it depends on the sum of bytes of all published-failed-messages or the number of all published-failed-messages, I guess).

So I looked into the source code of mosquitto, the flow of publishing is:

[exec mosquitto_publish()] --> [check the size of message] --> [if the size is valid, push_back to a mesages-linked-list].

In the net's loop, get the pointer of the first message from the list to publish, if publish success, remove it from list, and go to the next message; if failed, will not remove this message from this list.

You don't check the size of the current mesages-linked-list before exec push_back and the sum of bytes of all messages in this mesages-linked-list! This conflicted with my upper test. So, question one, why not to check?(if my upper analysis is right)

I want to add a feature into my client. If client publish failed with qos = 1 or 2, I will save them into local database, after network reworks, I'll republish and remove them from database manually.

But now it has some problems,

One: If publish failed with qos = 1, and after network reworks, both you and I will republish them, that is wasteful.

Two: If publish failed with qos = 2, and after network reworks, both you and I will republish them, this conflict with the definition of qos = 2.

So, my question two, is there a solution to stop your republish-action? (From my point, by removing from linked-list rather than still leaving them in linked-list when publish failed will solve this problem, and at the same time, this will be more friendly for developers.)

[ghost]

@ethsonliu . Did you find any fix or workaround for this issue ?

[ethsonliu]

@vishal000 Sry, no.