backend: retry datasource requests for trivial network errors

[hagen1778]

See VictoriaMetrics/VictoriaMetrics#6712

In the issue above, request sent by Grafana failed because of EOF error. There could be many reasons why plugin faced the EOF error, but the sensible thing to do would be to retry such trivial network errors.

For the reference, other VM components like vmalert already retry requests for such errors.

[dmitryk-dk]

Hi @hagen1778 ! Should we use the exact mechanism in the Victorialogs data source?

[hagen1778]

Hi @hagen1778 ! Should we use the exact mechanism in the Victorialogs data source?

Yes

[hagen1778]

plugin should retry trivial network errors like "unexpected EOF" or "broken pipe" in #194. This change will be included into the next release.

[hagen1778]

@dmitryk-dk can we make release soon?

[dmitryk-dk]

@dmitryk-dk can we make release soon?

Yeah, i think will do it today little bit later

[dmitryk-dk]

Hi @hagen1778 ! Release was published with the latest changes. The fix was included into v0.9.0 release.

[dmitryk-dk]

Hi @hagen1778 ! The same behavior was ported to the victorialogs-datasource in the v0.4.0 release