Queue names being truncated

[eberquist]

We are seeing documents labeled with queue names that are being truncated. We have a queue named low and while we do see documents with that queue label we also see them labeled as lo. This happens to other queues as well. We have verified that we indeed don't have any Sidekiq queues running under those names on the servers. Here is an example document that contains the incorrect queue name as lo. The issue is not contained to a single set of metrics either, we have queue_latency and job_latency metrics with the same issue.

{ "_index": "metricbeat-7.6.2-2022.06.17-000034", "_id": "dsl68oEBK-QS_yAzReSg", "_version": 1, "_score": 1, "_source": { "@timestamp": "2022-07-12T12:55:08.053Z", "labels": { "queue": "lo" }, "event": { "dataset": "statsd", "module": "statsd" }, "service": { "type": "statsd" }, "statsd": { "sidekiq_running_job_runtime_seconds": { "value": 0 } }, "fields": { "application": "", "env": "production" }, "ecs": { "version": "1.4.0" }, "metricset": { "name": "server" }, "host": { "name": "process1", "hostname": "process1", "architecture": "x86_64", "os": { "codename": "bionic", "platform": "ubuntu", "version": "18.04.5 LTS (Bionic Beaver)", "family": "debian", "name": "Ubuntu", "kernel": "5.4.0-1064-aws" }, "id": "", "containerized": false }, "agent": { "version": "7.6.2", "type": "metricbeat", "ephemeral_id": "", "hostname": "process1", "id": "" }, "cloud": { "instance": { "id": "" }, "machine": { "type": "" }, "region": "", "availability_zone": "", "account": { "id": "" }, "image": { "id": "" }, "provider": "" } }, "fields": { "host.os.name.text": [ "Ubuntu" ], "statsd.sidekiq_running_job_runtime_seconds.value": [ 0 ], "host.hostname": [ "process1" ], "cloud.availability_zone": [ "" ], "service.type": [ "statsd" ], "cloud.instance.id": [ "" ], "agent.type": [ "metricbeat" ], "event.module": [ "statsd" ], "host.os.version": [ "18.04.5 LTS (Bionic Beaver)" ], "host.os.kernel": [ "5.4.0-1064-aws" ], "host.os.name": [ "Ubuntu" ], "host.name": [ "process1" ], "host.id": [ "" ], "fields.env": [ "production" ], "cloud.region": [ "" ], "host.os.codename": [ "bionic" ], "agent.hostname": [ "process1" ], "fields.application": [ "" ], "cloud.image.id": [ "" ], "metricset.name": [ "server" ], "host.architecture": [ "x86_64" ], "cloud.provider": [ "" ], "labels.queue": [ "lo" ], "cloud.machine.type": [ "" ], "@timestamp": [ "2022-07-12T12:55:08.053Z" ], "agent.id": [ "" ], "host.containerized": [ false ], "ecs.version": [ "1.4.0" ], "cloud.account.id": [ "" ], "host.os.platform": [ "ubuntu" ], "agent.ephemeral_id": [ "" ], "agent.version": [ "7.6.2" ], "host.os.family": [ "debian" ], "event.dataset": [ "statsd" ] } }

[Envek]

This is super weird!

I suppose that you're using yabeda-statsd for exporting metrics, right?

I can't see anything suspicious neither in yabeda-statsd nor in yabeda-sidekiq nor in yabeda itself. AFAIK we don't do any fancy string operations that could lead to this and, honestly, have no idea how this might happen.

Don't know much about statsd though.