search

influxdata / docs-v2

InfluxData Documentation that covers InfluxDB Cloud, InfluxDB OSS 2.x, InfluxDB OSS 1.x, InfluxDB Enterprise, Telegraf, Chronograf, Kapacitor, and Flux.
https://docs.influxdata.com
MIT License
73 stars 275 forks source link

Influx Inspect Build TSI #1478

Open dp1140a opened 4 years ago

dp1140a commented 4 years ago

There should probably be some mention that the build tsi command needs to be run on all the data nodes. For that fact this may apply to several of the other subcommands on this page.

URL for relevant page?

https://docs.influxdata.com/influxdb/v1.8/tools/influx_inspect/ https://docs.influxdata.com/influxdb/v1.8/tools/influx_inspect/#buildtsi

Where did you look before opening the issue?

https://docs.influxdata.com/influxdb/v1.8/tools/influx_inspect/ https://docs.influxdata.com/influxdb/v1.8/tools/influx_inspect/#buildtsi

timhallinflux commented 4 years ago

Thanks @dp1140a

We had a similar issue come-up earlier this week with another user.

This is one of a number of issues related to having the OSS documentation and tooling attempting to cover for the InfluxDB Enterprise version as well.

We need to have some highlights in this docs which says something to the effect: For InfluxDB Enterprise, you must run this on each data node. Then...the steps: run influxd-ctl show. (Show output which lists data nodes). Start with the first data node listed. Go to that host and stop the influxd process on node. Run the buildtsi command.... when completed, restart the influxd process. Then...monitoring! Use the dashboards (pointer to platform docs for how to monitor InfluxDB Enterprise) and checkout the hinted-handoff graph. Wait until the hinted-handoff drains before proceeding to the next node....repeat the process until all nodes have been covered. (this is really similar to the rolling upgrade process works when you change a configuration and need to duplicate that change across the cluster...you must do this one node at a time, if you wish to avoid downtime).

There is another note in this same section: Note: For offline conversion only. -- this means very different things for OSS vs. Enterprise. So we should probably specify what this means. It means that the influxd process MUST NOT be running on the data node where you are attempting to run influx_inspect buildtsi. For OSS, there is only one data node, but for Enterprise, you can halt the influxd process on one data node, run influx_inspect buildtsi and then restart influxd before proceeding to the subsequent data nodes....