Mosca Crashing Server

[ccravens]

It appears that having the Mosca library included in the server may be crashing it. At, what appears to be random intervals, we get the following:

/Users/ccravens/Business/company/Clients/project/src/project-core/node_modules/ascoltatori/node_modules/mongodb/lib/utils.js:98
    process.nextTick(function() { throw err; });
                                  ^

Error: Cannot stat if collection is capped or not
    at /Users/ccravens/Business/company/Clients/project/src/project-core/node_modules/ascoltatori/lib/mongo_ascoltatore.js:235:26
    at handleCallback (/Users/ccravens/Business/company/Clients/project/src/project-core/node_modules/ascoltatori/node_modules/mongodb/lib/utils.js:96:12)
    at /Users/ccravens/Business/company/Clients/project/src/project-core/node_modules/ascoltatori/node_modules/mongodb/lib/collection.js:1503:20

[ccravens]

Actually this might be due to the connection to the database being poor. Might not be an ascoltatori issue at all... Just wanted to verify everyone's thoughts before closing it.

[mcollina]

I think it's a combination of both. The connection is poor, and ascoltatori is failing too early: https://github.com/mcollina/ascoltatori/blob/master/lib/mongo_ascoltatore.js#L232-L237.

I think we should add a retry mechanism there.

[dushyantbangal]

@mcollina has there been any fix for this? can you suggest a workaround? My MQTT just stops working after this error :(

[mcollina]

No I have not worked on a fix. Would you like to send a PR?

[dushyantbangal]

I've come up with the following (dirty) retry logic:

if (err) 
{
    if(!retryCount)
        retryCount=0;
    if(retryCount<that._maxRetry)
    {
        debug('checkCappedAndPoll -> Cannot stat isCapped. Retry');

        setTimeout(function(){
            that._checkCappedAndPoll(latest, ++retryCount);
        },500)

        return;
    }else{
        debug('checkCappedAndPoll -> Cannot stat isCapped. Give up');

        that.emit('error', new Error('Cannot stat if collection is capped or not'));
        that._handlingCursorFailure = false;
        return;
    }
}

Should I be doing it like this? Should I be putting the setTimeout of 500ms or directly retrying or putting larger timeout? Should I be using something else than your that._maxRetry ?

[mcollina]

@dushyantbangal does that solves your problem?

I think we can increase the maximum retry count to 5 and do something like setTimeout(func, 100 * retryCount), so that it increase overtime (and the first one is 0).

[dushyantbangal]

I'm not really able to recreate the issue, it happens sometimes. But I did try the following:

1. Set error=true for count 0,1,2
2. Execute the broker with DEBUG=ascoltatori:mongodb

For counts 0,1,2 the execution went in retry, and after count 2 the polls were successful.

[mcollina]

Can you send a PR then?

[dushyantbangal]

Sure, will do!