Specific plugin error message is not propagated to the client

[grinapo]

While testing #22 I created some not quite valid queries it seems, which is fine (or at least I believe that at this point :-)).
However all I got was a generic "500 An error occurred within the plugin".

Checking the grafana logs I see the real message there:

logger=context userId=2 orgId=1 uname=grin t=2024-06-26T15:46:57.572889164+02:00 level=error msg="Internal server error" error="[plugin.downstreamError] client: failed to query data: Failed to query data: rpc error: code = Unknown desc = frame has different field lengths, field 0 is len 360 but field 1 is len 0" remote_addr=xyz traceID=

I am not sure that it's possible to forward the specific message to the client (most plugins do that), but if it is, it would be extremely useful.

ps: "severity:info | stats by (_time:1m) count() logs"

[dmitryk-dk]

Hi @grinapo ! Could you please try to upgrade the plugin to the latest release and be sure that the backend part or Grafana was restarted.? Why I am asking - this issue was fixed in the newest release, and I have checked your query on my local and our prod end, and it works well; just shows me no data message. But do not return any error.

[crocchietta]

I have version 0.2.2 (yesterday) installed.
severity:info | stats by (_time:1m) count() logs also doesn't generate an error. But if there is a result, it generates the following error:
logger=context userId=1 orgId=1 uname=admin_grafana t=2024-06-26T16:17:34.641025924+02:00 level=error msg="Internal server error" error="[plugin.downstreamError] client: failed to query data: Failed to query data: rpc error: code = Unknown desc = frame has different field lengths, field 0 is len 15 but field 1 is len 0" remote_addr=w.x.y.z traceID=

query in victorialogs vmui :
_stream:{job="fail2ban_log"} service:spb | stats by (_time:1m) count() logs

[
  {
    "_time": "2024-06-26T14:11:00Z",
    "logs": "2"
  },
  {
    "_time": "2024-06-26T14:17:00Z",
    "logs": "1"
  }
]

[dmitryk-dk]

@grinapo , @crocchietta ! Thank you for the provided information. I found the reason and prepared the fix. Need to test it a little bit

[crocchietta]

When is the next release with this fix planned?

[dmitryk-dk]

When is the next release with this fix planned?

Maybe i will do it today

[crocchietta]

Thanks.

[crocchietta]

I just tried the fix and there are no more errors.

On the other hand, I can't use the result of the query with time series visualization. I get a "Data is missing a number field".
Here is the result of the query:

2024-06-28 09:50:00.000	{"logs":"2"}
2024-06-28 07:32:00.000	{"logs":"1"}
2024-06-28 07:35:00.000	{"logs":"2"}
2024-06-28 08:04:00.000	{"logs":"1"}
2024-06-28 08:35:00.000	{"logs":"2"}

But maybe it's not possible to do that ?

[dmitryk-dk]

Hi @grinapo , @crocchietta ! Here is the release. Could you check it, please? If it works correctly, can you close the issue?

[crocchietta]

It's ok for me.

@dmitryk-dk : Is it possible to use the result of a query for serial times (as with Loki) or is it only planned for later?

[grinapo]

@dmitryk-dk this issue seems to have been taken over by the evil stream field id problem (which indeed have been fixed), but the original issue was more generic: the plugin shall hand over to grafana the specific error message, and not just "got unexpected response status code: 400".

(to test: foo | stats evilverb (_time:1m) count() - should tell me where and what the error was.)

[dmitryk-dk]

Hi @grinapo , @crocchietta ! We can close this issue related to the described problem. And you can open new ones with improvements to the errors. If this works for you, I would appreciate a new issue from your side, and I'll close this one.

Thank you:)

[crocchietta]

it's ok for me.

[dmitryk-dk]

Close as solved. Open new issue related error improvements, please. Thank you!