Bull: removeJobs function returns redis error

[DavideViolante]

Description

Not sure if Bull bug or something else.
I'm trying to call queue.removeJobs(pattern) with pattern = "12345*" but it returns:

ERR Error running script (call to f_531d166ed840471540203603686123486730c4d5): @user_script:38: @user_script: 38: Write commands not allowed after non deterministic commands

Bull version

v3.14.0

[manast]

hmm, that strange, are you running in a master-slave setup?

[DavideViolante]

No

[manast]

can you provide a complete example that reproduces the issue?

[james-siteclick]

I get the same issue. Here is an example script:

require('dotenv').config();
const Bull = require('bull');
const uuid = require('uuid/v4');

const isLocal = process.env.NODE_ENV === 'local';

const config = {
  redis: {
    host: process.env.REDIS_HOST || '127.0.0.1',
    port: process.env.REDIS_PORT || 6379,
    password: process.env.REDIS_PASSWORD || '',
    ...!isLocal && { tls: {} },
    db: 2
  }
}

const queue = new Bull('test-queue', config);

const processor = async data => {
  console.log('processing job...'), data;
  throw new Error('Oh no!');
}

queue.process(processor);

queue.on('failed', async () => {
  await queue.removeJobs('test-queue*');
  console.log('queue deleted OK');
});

queue.add({ foo: 'bar' }, {
  jobId: `test-queue.${uuid()}`
});

On my local machine this works, however on AWS with Elasticache I get the error described above.

[DavideViolante]

We are also using AWS Elasticache.

[denysaw]

Same thing here :(

ReplyError: ERR Error running script (call to f_531d166ed840471540203603686123486730c4d5): @user_script:38: @user_script: 38: Write commands not allowed after non deterministic commands

Error on removeJobs, not on getJob for sure. Job is completed, not processing.

Redis is used ONLY for Bull.

[denysaw]

Oh, easily removed with job.remove(). So I don't need any help, but if you wish you still could solve this issue.

[manast]

Which version of elasticache are you using? I use remove with pattern in elasticache without any issues so far.

[manast]

@james-siteclick is your intent to delete all the jobs in a queue with the above pattern? Can you try with

 await queue.removeJobs('*');

instead? you do not need to specify the "prefix" of the queue.

[manast]

@denysaw did you get the above error locally also or only on elasticache?

[denysaw]

@denysaw did you get the above error locally also or only on elasticache?

Locally. Today also tried to remove all jobs on start (as empty() does nothing): await queue.removeJobs('*'); getting same error.

[denysaw]

Thank God, I thought of such workaround:

for (let job of await queue.getJobs()) {
    await job.moveToFailed({message: 'Canceled'}, true);
    await job.remove();
}

Worx. But I would be glad to replace this *code with working empty() or removeJobs() ;)

[manast]

@denysaw empty works just fine, removes all the jobs that are waiting or delayed.

[manast]

@denysaw whats your version of redis?

[denysaw]

Strange. I started today with empty(), but getJobs() shown me job I started yesterday :)
Redis is 3.0.504

[manast]

Can you try with redis 5.0.3 instead? 3.x is pretty old actually, could be an issue with redis itself.

[denysaw]

Can't right now. Will try in some days only. Gonna let you know ;)

[james-siteclick]

@james-siteclick is your intent to delete all the jobs in a queue with the above pattern? Can you try with

 await queue.removeJobs('*');

instead? you do not need to specify the "prefix" of the queue.

@manast thanks for your reply. In our use case we have a parent job with an ID something like:

And this creates some other jobs:

parent-job-id.child-job-id-a
parent-job-id.child-job-id-b
...

We are trying to delete only the jobs that were created by this particular parent job, so we are doing something like:

queue.removeJobs('parent-job-id.*');

I simplified this a bit for the code example.

[manast]

@james-siteclick that should work. Could you just verify that your elasticache in AWS is running version 5.x ? I have services in production where we use remove with pattern in AWS and it is working, so I am a bit puzzled by this. The error itself does not make so much sense, since the call is deterministic.

[rhsobr]

Can you try with redis 5.0.3 instead? 3.x is pretty old actually, could be an issue with redis itself.

This error is gone after updating Redis version to 5.x.

[IRTip]

Getting the same error. Seems like this error happens when a pattern matches a job.
Versions:

OS: Win10
Redis: 3.0.2
Bull: 3.18.0

You can replicate the error with the following code:

(async () => {
await queue.add({}, {jobId: 'a'});
await queue.removeJobs('a');
)();