[0.10.0] panic: runtime error: index out of range when executing continuous query

[malaakso]

I yesterday upgraded from 0.9.6 to 0.10.0 (did not convert old data to tsm), and added a new CQ. At midnight InfluxDB paniced:

[continuous_querier] 2016/02/18 00:00:00 executing continuous query r (2016-02-17 00:00:00 +0000 UTC to 2016-02-18 00:00:00 +0000 UTC)
[query] 2016/02/18 00:00:00 SELECT count(a) + count(b) INTO s."default".c FROM s."default".d WHERE time >= '2016-02-17T00:00:00Z' AND time < '2016-02-18T00:00:00Z' GROUP BY time(1d), SIN
panic: runtime error: index out of range

goroutine 135548 [running]:
github.com/influxdb/influxdb/models.scanTo(0xc209367f78, 0x2, 0x8, 0x0, 0xc20935a03d, 0x41d413, 0xc20935a008, 0x410868, 0x7f87086eee60)
        /tmp/tmp.r1Nea5VXEw/src/github.com/influxdb/influxdb/models/points.go:892 +0xad
github.com/influxdb/influxdb/models.newFieldsFromBinary(0xc209367f78, 0x2, 0x8, 0xc20935a040)
        /tmp/tmp.r1Nea5VXEw/src/github.com/influxdb/influxdb/models/points.go:1396 +0x10d
github.com/influxdb/influxdb/models.(*point).unmarshalBinary(0xc20829b050, 0xc2090bf6e0)
        /tmp/tmp.r1Nea5VXEw/src/github.com/influxdb/influxdb/models/points.go:1298 +0x3b
github.com/influxdb/influxdb/models.(*point).Fields(0xc20829b050, 0xc2094810b0)
        /tmp/tmp.r1Nea5VXEw/src/github.com/influxdb/influxdb/models/points.go:1211 +0x41
github.com/influxdb/influxdb/tsdb.(*Shard).validateSeriesAndFields(0xc2082fb760, 0xc209367fc0, 0x1, 0x1, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, ...)
        /tmp/tmp.r1Nea5VXEw/src/github.com/influxdb/influxdb/tsdb/shard.go:413 +0x7e5
github.com/influxdb/influxdb/tsdb.(*Shard).WritePoints(0xc2082fb760, 0xc209367fc0, 0x1, 0x1, 0x0, 0x0)
        /tmp/tmp.r1Nea5VXEw/src/github.com/influxdb/influxdb/tsdb/shard.go:202 +0xa6
github.com/influxdb/influxdb/tsdb.(*Store).WriteToShard(0xc208088160, 0xbc, 0xc209367fc0, 0x1, 0x1, 0x0, 0x0)
        /tmp/tmp.r1Nea5VXEw/src/github.com/influxdb/influxdb/tsdb/store.go:442 +0x14e
github.com/influxdb/influxdb/cluster.func<C2><B7>002(0xbc, 0x1, 0xc209367fc0, 0x1, 0x1)
        /tmp/tmp.r1Nea5VXEw/src/github.com/influxdb/influxdb/cluster/points_writer.go:316 +0x103
created by github.com/influxdb/influxdb/cluster.(*PointsWriter).writeToShard
        /tmp/tmp.r1Nea5VXEw/src/github.com/influxdb/influxdb/cluster/points_writer.go:352 +0x31a

[jonseymour]

G'day @malaakso.

This looks like a duplicate of #5664.

This issue has been partially fixed by #5697 (which is in master). Depending on what is causing it in your particular case, you may also need one or more of #5732 (being developed by the influx guys). My fix #5716 also addresses this issue in a different way.

To take advantages of these fixes right now you need to create a local build.

The issue is caused in part by empty column names that are currently created when you do maths on an aggregate with a SELECT INTO query, something like:

select count(value)/100 into "downsample" time > now() - 4d group by time(1h)

A workaround is to add aliases to all columns which do maths of SELECT INTO statements in your CQs, in this case:

select count(value)/100 as sampled into "downsample" time > now() - 4d group by time(1h)

The workaround may be the easiest thing to try at this point.

[malaakso]

Thanks, I'll add an alias to the CQ.

[e-dard]

Closing as dupe of #5664.

@malaakso as @jonseymour says, adding an alias is a good work around at the moment.