LIMIT returns a non-deterministic series

[beckettsean]

I'm using telegraf to write data, it writes a series for each CPU core, so I have 8 tag sets underneath the cpu_system measurement. Here's the most recent point from all 8 series:

> select value, cpu from cpu_system limit 1
name: cpu_system
tags: cpu=cpu0
time                value
----                -----
2015-06-26T19:21:54.661779022Z  61665.6171875


name: cpu_system
tags: cpu=cpu1
time                value
----                -----
2015-06-26T19:21:54.661779022Z  4253.4453125


name: cpu_system
tags: cpu=cpu2
time                value
----                -----
2015-06-26T19:21:54.661779022Z  34324.8046875


name: cpu_system
tags: cpu=cpu3
time                value
----                -----
2015-06-26T19:21:54.661779022Z  4455.140625


name: cpu_system
tags: cpu=cpu4
time                value
----                -----
2015-06-26T19:21:54.661779022Z  35394.4921875


name: cpu_system
tags: cpu=cpu5
time                value
----                -----
2015-06-26T19:21:54.661779022Z  4444.15625


name: cpu_system
tags: cpu=cpu6
time                value
----                -----
2015-06-26T19:21:54.661779022Z  34706.5703125


name: cpu_system
tags: cpu=cpu7
time                value
----                -----
2015-06-26T19:21:54.661779022Z  4447.0546875

If I then run the select statement without selecting the tag (cpu) then I get the most recent point from a randomly selected series:

(note that the timestamps are identical, and the values are all contained in the set of 8 above, and they do repeat randomly)

@pauldix and I agree this should be deterministic behavior. According to him, LIMIT should pull from the series with the lowest index. Since I'm on a single-node system the index should be consistent. However, the series chosen is not.

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  61665.6171875

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  34324.8046875

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  34324.8046875

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  35394.4921875

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  4447.0546875

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  34706.5703125

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  61665.6171875

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  4447.0546875

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  61665.6171875

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  4444.15625

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  4253.4453125

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  4447.0546875

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  34706.5703125

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  4447.0546875

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  4444.15625

> select value from cpu_system limit 1
name: cpu_system
----------------
time                value
2015-06-26T19:21:54.661779022Z  4253.4453125

[jsternberg]

@beckettsean is it possible to check if this is still valid with the new query engine?

[beckettsean]

It is still not deterministic in 0.11.1:

> select * from cpu limit 1
name: cpu
---------
time            cpu host        usage_guest usage_guest_nice    usage_idle      usage_iowait    usage_irq   usage_nice  usage_softirq   usage_steal usage_system        usage_user
1456350580000000000 cpu1    sean-stable 0       0           99.79939819458393   0       0       0       0       0       0.10030090270812698 0.10030090270812253

> select * from cpu limit 1
name: cpu
---------
time            cpu host        usage_guest usage_guest_nice    usage_idle      usage_iowait    usage_irq   usage_nice  usage_softirq   usage_steal usage_system        usage_user
1456350580000000000 cpu0    sean-stable 0       0           99.49899799599173   0       0       0       0       0       0.30060120240480265 0.20040080160320178

[beckettsean]

explicit ORDER BY does not introduce determinism, either:

> select * from cpu order by time asc limit 1
name: cpu
---------
time            cpu host        usage_guest usage_guest_nice    usage_idle      usage_iowait    usage_irq   usage_nice  usage_softirq   usage_steal usage_system        usage_user
1456350580000000000 cpu0    sean-stable 0       0           99.49899799599173   0       0       0       0       0       0.30060120240480265 0.20040080160320178

> select * from cpu order by time asc limit 1
name: cpu
---------
time            cpu     host        usage_guest usage_guest_nice    usage_idle      usage_iowait    usage_irq   usage_nice  usage_softirq       usage_steal usage_system        usage_user
1456350580000000000 cpu-total   sean-stable 0       0           99.59919839679381   0       0       0       0.05010020040080163 0       0.2004008016032068  0.15030060120241065

[jsternberg]

Ok, I'll track this as something to potentially work on for 0.13.