Results of queries with that involve the same field key panics

[desa]

Built from source v0.9.x

The query

select value * value from cpu_user limit 1

yields two value columns

name: cpu_user
--------------
time                value   value
2015-06-26T21:34:26.85363181Z   1.4758622874084473e+07

Similarly the query

select value * value * value from cpu_user limit 1

yields tree value columns.

name: cpu_user
--------------
time                value   value   value
2015-06-26T21:34:26.85363181Z   2.7929823501472473e+06

[beckettsean]

@corylanou another one for the columns returned from queries

[otoolep]

Spoke with @mjdesa briefly about this, it's a weird one. Basically the query-engine code sees two "fields" in the query, and therefore preps two columns in the output. It might be a quick fix, it might not be.

[otoolep]

Queries such as these now cause the server to panic.

[pauldix]

@otoolep can you confirm if this still panics the server?

[otoolep]

I believe it does still panic the system, but I got side tracked by other
work. Let me get back into this.

On Tuesday, August 18, 2015, Paul Dix [email protected] wrote:

@otoolep https://github.com/otoolep can you confirm if this still
panics the server?

—
Reply to this email directly or view it on GitHub
#3200 (comment).

[beckettsean]

It does still panic the system.

root@sean-test:~# /opt/influxdb/influx
Connected to http://localhost:8086 version 0.9.3-nightly-1548f62
InfluxDB shell 0.9.3-nightly-1548f62
> select value * value from thing
ERR: Get http://localhost:8086/query?db=foo&q=select+value+%2A+value+from+thing: EOF
panic: runtime error: index out of range

First few lines of the panic:

goroutine 148 [running]:
github.com/influxdb/influxdb/influxql.func·016(0xc2080dc960, 0x2, 0x2, 0x8fbe20, 0xc208083f30)
        /tmp/tmp.yrqIWEIVNQ/src/github.com/influxdb/influxdb/influxql/result.go:146 +0x4a
github.com/influxdb/influxdb/influxql.func·020(0xc2080dc960, 0x2, 0x2, 0x0, 0x0)
        /tmp/tmp.yrqIWEIVNQ/src/github.com/influxdb/influxdb/influxql/result.go:186 +0xa8
github.com/influxdb/influxdb/tsdb.processForMath(0xc208038130, 0x1, 0x1, 0xc2081197d0, 0x2, 0x2, 0x0, 0x0, 0x0)
        /tmp/tmp.yrqIWEIVNQ/src/github.com/influxdb/influxdb/tsdb/executor.go:867 +0x4ad
github.com/influxdb/influxdb/tsdb.(*limitedRowWriter).processValues(0xc20809d2c0, 0xc20813a000, 0x2, 0x2710, 0x0)
        /tmp/tmp.yrqIWEIVNQ/src/github.com/influxdb/influxdb/tsdb/executor.go:760 +0xf45
github.com/influxdb/influxdb/tsdb.(*limitedRowWriter).Flush(0xc20809d2c0)
        /tmp/tmp.yrqIWEIVNQ/src/github.com/influxdb/influxdb/tsdb/executor.go:661 +0xd6
github.com/influxdb/influxdb/tsdb.(*Executor).executeRaw(0xc2081187e0, 0xc2080f50e0)
        /tmp/tmp.yrqIWEIVNQ/src/github.com/influxdb/influxdb/tsdb/executor.go:228 +0x4e2
created by github.com/influxdb/influxdb/tsdb.(*Executor).Execute
        /tmp/tmp.yrqIWEIVNQ/src/github.com/influxdb/influxdb/tsdb/executor.go:88 +0x9a

[otoolep]

Yeah, I'm in this code now.

This may not be a quick fix, there is some confused code (the code is confused, not me :-) ) within our system and I'm trying to see what is the best way to resolve this.

[otoolep]

Changes to fix this will potentially ripple through the code, and it may be even more complicated to fix queries such as value 1* value2 where the data point fields within the query are actually different (but the parser considers it a single field).

It may be too close to the 0.9.3 freeze to consider the changes this would have on the code base.

[otoolep]

Pushing out to 0.9.4.

[otoolep]

Too much else going on in 0.9.4 - pushing out.

[beckettsean]

closing in favor of #4046, which is more generic and cleaner.