The number of writer goroutines is defined by the write-concurrency configuration. setting the subscriber write-concurrency to greater than 1 increases subscriber write throughput, it can result in out-of-order writes under high ingest rates. Setting write-concurrency to 1 ensures writes are passed to subscriber endpoints sequentially, but can create a bottleneck under high ingest rates. InfluxQL subscription statements
SHOW SUBSCRIPTIONS | ||||||||
example influxdb.conf subscriber configuration:
[subscriber] enabled = true http-timeout = "30s" insecure-skip-verify = false ca-certs = "" write-concurrency = 40 write-buffer-size = 1000Descriptions of [subscriber] configuration options are available in the Configuring InfluxDB documentation.
"Post http://x.y.z.a:9092/write?consistency=...: net/http: request canceled while waiting for connection (Client.Timeout exceeded while awaiting headers)" ... service=subscriber "Post http://x.y.z.a:9092/write?consistency=...: dial tcp x.y.z.a:9092: getsockopt: connection refused" ... service=subscriber "Post http://x.y.z.a:9092/write?consistency=...: dial tcp 172.31.36.5:9092: getsockopt: no route to host" ... service=subscriberThis may be caused by a networking error preventing a successful connection to the subscription endpoint. or the subscription endpoint no longer exists and the subscription hasn't been dropped.
InfluxDB does not know if a subscription endpoint will become accessible again, subscriptions are not automatically dropped when an endpoint becomes inaccessible.
Log entries like:
2019-12-05 16:31:12 daemon.info influxd[15135]: ts=2019-12-05T21:31:12.382670Z lvl=info
msg="Post http://localhost:9092/write?consistency=&db=_internal&precision=ns&rp=monitor:
dial tcp [::1]:9092: connect: connection refused" log_id=0JXX36AG000 service=subscriber
Are due to subscription
Edit /etc/influxdb/influxdb.conf and set subscriotion enable = false