Amazing product. We use VictoriaMetrics for quite a while, and previously used Loki and our custom Clickhouse/Vector approach for logs and we have switched to VictoriaLogs. It is much better and faster than Loki, same goes with custom CH/Vector thing we had. Kudos to the team, we are waiting for VictoriaTraces to switch Tempo instance to it for opentelemetry stuff.

▲

CubsFan1060 4 days ago | parent [-]

Can you talk a little bit about your Victoria Logs setup? About how many logs are you ingesting and what kind of sizing do you have on your setup?

▲

nklmilojevic 3 days ago | parent | next [-]

Sure thing!

Ingested logs 24h: 428 Mil Ingested bytes 24h: 625GB Inser req/s: 6k/s

8vCPU, 16GB mem. Running standard-rwo PVC on GCP.

We have a couple of projects like this with similar usage and similar machine sizing.

Still running vmlogs-single, and we will until we see a need to move to vmlogs-cluster version.

▲

camel_gopher 3 days ago | parent | next [-]

That sounds like a lot of resources provisioned for 6k/s. 625GB/24hr is a small footprint.

▲

CubsFan1060 3 days ago | parent [-]

I would have said it sounded pretty good. What technologies are you comparing it against, out of curiosity.

▲

camel_gopher 3 days ago | parent [-]

ClickHouse

	▲	CubsFan1060 3 days ago \| parent [-]
		Can you share any additional details? What kind of ingestion do you have, with what dimensions of a clickhouse cluster? I'm also curious how it handles structured vs unstructured logs. Thanks!

▲

CubsFan1060 3 days ago | parent | prev [-]

That seems pretty good. Do you have any sort of HA solution?

▲

PaywallBuster 3 days ago | parent | prev [-]

personally Hetzner SX295 that has 14x 22 TB on a ZFS setup

It ingests 70k lines per second without a sweat

reads are just as fast