Hm. Anybody have any idea what an error like `[ha...
# support
e
Hm. Anybody have any idea what an error like
[handshake] unexpected packet [72] from server
from the clickhouse exporters might indicate? (logs threaded)
Copy code
Oct 09 21:11:25 durian000 signoz-otel-collector[1086820]: 2023-10-09T21:11:25.743Z        info        service/telemetry.go:104        Setting up own telemetry...
Oct 09 21:11:25 durian000 signoz-otel-collector[1086820]: 2023-10-09T21:11:25.743Z        info        service/telemetry.go:127        Serving Prometheus metrics        {"address": "127.0.0.1:8888", "level": "Basic"}
Oct 09 21:11:25 durian000 signoz-otel-collector[1086820]: 2023-10-09T21:11:25.743Z        info        exporter@v0.79.0/exporter.go:275        Stability level of component is undefined        {"kind": "exporter", "data_type": "traces", "name": "clickhousetraces"}
Oct 09 21:11:25 durian000 signoz-otel-collector[1086820]: Error: failed to build pipelines: failed to create "clickhousetraces" exporter for data type "traces": error connecting to primary db: [handshake] unexpected packet [72] from server
Oct 09 21:11:25 durian000 signoz-otel-collector[1086820]: 2023/10/09 21:11:25 application run finished with error: failed to build pipelines: failed to create "clickhousetraces" exporter for data type "traces": error connecting to primary db: [handshake] unexpected packet [72] from server
Oct 09 21:11:25 durian000 systemd[1]: signoz-otel-collector.service: Main process exited, code=exited, status=1/FAILURE
(This is a relatively custom install on bare metal; no Docker or anything like that. Still figuring out all of the config and whatnot, it’s probably some really basic user-error.)
s
What is the DSN configured? The clickhouse exporters in SigNoz only support tcp. It’s probably an issue because of mis configuration.
e
Ah, figured out that Clickhouse’s HTTP protocol isn’t implemented shortly after I posted the above (oof, I hate trying to use multiple Slacks) — tell-tale
unexpected packet [72]
, since ASCII 72 is
H
, in
HTTP/1…
etc. Is HTTP-protocol support for Clickhouse on the radar? Or alternatively, can I configure multiple host/port combinations and distribute across them? Talking to only a single server is something of a non-starter for us … HTTP allows us to use our usual NginX tunneling/balancing infrastructure for this, which is preferable. A custom streaming Clickhouse-protocol balancer would be a significant amount of engineering effort. Hrm. :x