On 06/05/2015 04:02 PM, Thomas Segismont wrote: > Hi everyone, > > While doing some testing for Grafana/Metrics integration, I found that > if you don't supply a timestamp when you send data points to Metrics, > they all get stored with timesptamp=0. This not right. > > In such situations, Influx saves the point with the server timestamp. We > could also reply with 400 Bad Request. My preference goes to the latter. Agreed that reporting that it's wrong seems better that silently use some other value. But would we still have a simple way to explicitly use server timestamp ? Might be useful when you are unsure that the client machine has the right time and not care about some small time difference.

Thomas > > Any thoughts? > > I'll file a JIRA when we agree on a solution. > > Regards, > Thomas > _______________________________________________ > hawkular-dev mailing list > hawkular-dev(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/hawkular-dev > _______________________________________________ hawkular-dev mailing list hawkular-dev(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/hawkular-dev

Le 05/06/2015 16:12, Thomas Heute a écrit : > > > On 06/05/2015 04:02 PM, Thomas Segismont wrote: >> Hi everyone, >> >> While doing some testing for Grafana/Metrics integration, I found that >> if you don't supply a timestamp when you send data points to Metrics, >> they all get stored with timesptamp=0. This not right. >> >> In such situations, Influx saves the point with the server timestamp. We >> could also reply with 400 Bad Request. My preference goes to the latter. > > Agreed that reporting that it's wrong seems better that silently use > some other value. > > But would we still have a simple way to explicitly use server timestamp ? > Might be useful when you are unsure that the client machine has the > right time and not care about some small time difference. > Adding a flag (-Dhawkular.metrics.useServerTimestamp) to switch between failing requests and using a server timestamp would do the trick. In any case it's important we insist that users should make sure all machines times are synchronized.

> Thomas > >> >> Any thoughts? >> >> I'll file a JIRA when we agree on a solution. >> >> Regards, >> Thomas >> _______________________________________________ >> hawkular-dev mailing list >> hawkular-dev(a)lists.jboss.org >> https://lists.jboss.org/mailman/listinfo/hawkular-dev >> > _______________________________________________ > hawkular-dev mailing list > hawkular-dev(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/hawkular-dev > _______________________________________________ hawkular-dev mailing list hawkular-dev(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/hawkular-dev