v1,0-beta1 GROUP BY returns multiple same timestamps?!

[ckdarby]

Bug report

System info:

Branch  Build Time  Commit                      Version
master          bf3c22689b68a6d03d20eac695a7049a7c78b21d    1.0.0-beta1

name: network
-------------
hostname
kappatron

name: runtime
-------------
GOARCH  GOMAXPROCS  GOOS    version
amd64   1       linux   go1.6.2

name: system
------------
PID currentTime         started             uptime
2787    2016-06-20T14:33:48.173023573Z  2016-06-18T21:20:54.006175158Z  41h12m54.166849308s

Steps to reproduce:

select count(value) from requests where time > now() - 3h GROUP BY time(15m)
name: requests
--------------
time            count
1466421300000000000 11212
1466422200000000000 45422
1466423100000000000 351
1466422200000000000 494
1466423100000000000 10902
1466422200000000000 428
1466423100000000000 8988
1466422200000000000 762
1466423100000000000 15185
1466422200000000000 214
1466423100000000000 13442
1466422200000000000 553
1466423100000000000 12090
1466424000000000000 27621
1466424900000000000 22248
1466425800000000000 34879
1466426700000000000 41229
1466427600000000000 45568
1466428500000000000 44769
1466429400000000000 45917
1466430300000000000 46072
1466431200000000000 11998
1466422200000000000 26
1466423100000000000 86
1466424000000000000 0
1466424900000000000 0
1466425800000000000 0
1466426700000000000 0
1466427600000000000 0
1466428500000000000 0
1466429400000000000 0
1466430300000000000 0
1466431200000000000 33484
1466432100000000000 27096

Expected behavior: Groupby returns a non-repeated timestamp

Actual behavior: Has duplicate timestamps

[ckdarby]

@jsternberg I just attempted with beta2 & this still happens

[jsternberg]

This might be related to #6738.

[jsternberg]

@ckdarby is there any chance you can make a tarball of your data, wal, and meta directories and upload it somewhere for me to check? If it's not possible for some reason, have you found any way to reproduce this consistently?

[ckdarby]

@jsternberg I ran with 0.13 and everything works fine now. #6738 was exactly what I was experiencing.

Unable to tarball due to the nature of data.

It was always reproducible. Different spots in the data. Send a total of about 200 million 1 value points across a span of 6 days. Data was sent to influxdb in non-ordered fashion. Send in batches of about 1 million at a time, followed by another 1 million shortly after in a repeating fashion.

These are the tags:
'endpoint' (string)
'domain' (string)
'status' (int)

For fields just value. Sending time with precision nanosecond.

[jsternberg]

Ok then. I'm going to close this as a duplicate of #6738. I think the problem still exists, but I'm not really sure at this point. Did you do anything to try and modify the database or did it just fix itself when you upgraded?

Thanks.

[ckdarby]

I purged the install, deleting everything, installed 0.13 and reimported
the same data without any issue.
On Jun 21, 2016 1:25 AM, "Jonathan A. Sternberg" notifications@github.com
wrote:

Closed #6877 #6877.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#6877 (comment), or mute
the thread
https://github.com/notifications/unsubscribe/AANce4ofbAljr27uW4WeBHXNYH79PIDKks5qN3WugaJpZM4I5w1e
.