We have been running influx 1.5.2 in production for about 6 months, we wanted to upgrade to influx 1.7.1. After upgrading the rpm, and restarting influx we now get logs similar to the following over and over. The same error is returned for many shards.
Dec 4 16:21:59 vpl-flux0001x influxd: ts=2018-12-04T22:21:59.134552Z lvl=info msg=“index opened with 8 partitions” log_id=0CApkQqG000 index=tsi
Dec 4 16:22:00 vpl-flux0001x influxd: ts=2018-12-04T22:22:00.548908Z lvl=warn msg=“error opening fields.idx: proto: invalid UTF-8 string. Rebuilding.” log_id=0CApkQqG000 engine=tsm1
Dec 4 16:22:00 vpl-flux0001x influxd: ts=2018-12-04T22:22:00.818915Z lvl=info msg=“Opened file” log_id=0CApkQqG000 engine=tsm1 service=filestore path=/data01/data/telegraf/autogen/634/000017075-000000004.tsm id=1 duration=211.981ms
I found the following one-liner to identify non-UTF-8 characters.
grep -axv ‘.*’ fields.idx
Running it returns a handful of unprintable characters but I get this result on other fields.idx files that influx isn’t complaining about.
I looked at the influx_inspect utility but didn’t see any options for looking at fields.idx.
I reviewed github issues and did some google searches but I couldn’t find anything. Anyone know of a way to recover from this? I would like to preserve my historical data.