fix(datadog_logs sink): retry HTTP requests and improve datadog sink error handling consistency #13130
Conversation
neuronull
added
sink: datadog_metrics
✅ Deploy Preview for vector-project ready!
To edit notification comments on pull requests, go to your Netlify site settings. |
github-actions
bot
added
the
domain: sinks
neuronull
requested review from
jszwedko,
fuchsnj,
spencergilbert and
StephenWakely
Soak Test ResultsBaseline: 8046713 ExplanationA soak test is an integrated performance test for vector in a repeatable rig, with varying configuration for vector. What follows is a statistical summary of a brief vector run for each configuration across SHAs given above. The goal of these tests are to determine, quickly, if vector performance is changed and to what degree by a pull request. Where appropriate units are scaled per-core. The table below, if present, lists those experiments that have experienced a statistically significant change in their throughput performance between baseline and comparision SHAs, with 90.0% confidence OR have been detected as newly erratic. Negative values mean that baseline is faster, positive comparison. Results that do not exhibit more than a ±8.87% change in mean throughput are discarded. An experiment is erratic if its coefficient of variation is greater than 0.3. The abbreviated table will be omitted if no interesting changes are observed. No interesting changes in throughput with confidence ≥ 90.00% and absolute Δ mean >= ±8.87%: Fine details of change detection per experiment.
|
|
Agree with sharing code. I'm also curious about this behavior across other HTTP-based sinks? We should be retrying these errors for all sinks. |
Agreed. We could take a similar tact as we have for the AWS components where they all have custom retry behavior but then fallback to shared retry logic: Lines 36 to 73 in f6f38d0 |
It looks like a lot of them (splunk_hec_metrics , splunk_hec_logs, clickhouse, influxdb, sematext, prometheus) use the generic Some others which don't use HttpRetryLogic:
That's not an exhaustive analysis but, it seems that many of the sinks retry all HttpErrors , rather than being selective, as was suggested in the comments here in this PR ... Perhaps |
Soak Test ResultsBaseline: 3210771 ExplanationA soak test is an integrated performance test for vector in a repeatable rig, with varying configuration for vector. What follows is a statistical summary of a brief vector run for each configuration across SHAs given above. The goal of these tests are to determine, quickly, if vector performance is changed and to what degree by a pull request. Where appropriate units are scaled per-core. The table below, if present, lists those experiments that have experienced a statistically significant change in their throughput performance between baseline and comparision SHAs, with 90.0% confidence OR have been detected as newly erratic. Negative values mean that baseline is faster, positive comparison. Results that do not exhibit more than a ±8.87% change in mean throughput are discarded. An experiment is erratic if its coefficient of variation is greater than 0.3. The abbreviated table will be omitted if no interesting changes are observed. No interesting changes in throughput with confidence ≥ 90.00% and absolute Δ mean >= ±8.87%: Fine details of change detection per experiment.
|
Sounds like outside the scope of this issue. Jesse has referenced this in #10870 . |
…:vectordotdev/vector into neuronull/datadog_logs_sink_http_retries
Soak Test ResultsBaseline: ce08bc8 ExplanationA soak test is an integrated performance test for vector in a repeatable rig, with varying configuration for vector. What follows is a statistical summary of a brief vector run for each configuration across SHAs given above. The goal of these tests are to determine, quickly, if vector performance is changed and to what degree by a pull request. Where appropriate units are scaled per-core. The table below, if present, lists those experiments that have experienced a statistically significant change in their throughput performance between baseline and comparision SHAs, with 90.0% confidence OR have been detected as newly erratic. Negative values mean that baseline is faster, positive comparison. Results that do not exhibit more than a ±8.87% change in mean throughput are discarded. An experiment is erratic if its coefficient of variation is greater than 0.3. The abbreviated table will be omitted if no interesting changes are observed. Changes in throughput with confidence ≥ 90.00% and absolute Δ mean >= ±8.87%:
Fine details of change detection per experiment.
|
…:vectordotdev/vector into neuronull/datadog_logs_sink_http_retries
Soak Test ResultsBaseline: 91d88e8 ExplanationA soak test is an integrated performance test for vector in a repeatable rig, with varying configuration for vector. What follows is a statistical summary of a brief vector run for each configuration across SHAs given above. The goal of these tests are to determine, quickly, if vector performance is changed and to what degree by a pull request. Where appropriate units are scaled per-core. The table below, if present, lists those experiments that have experienced a statistically significant change in their throughput performance between baseline and comparision SHAs, with 90.0% confidence OR have been detected as newly erratic. Negative values mean that baseline is faster, positive comparison. Results that do not exhibit more than a ±8.87% change in mean throughput are discarded. An experiment is erratic if its coefficient of variation is greater than 0.3. The abbreviated table will be omitted if no interesting changes are observed. No interesting changes in throughput with confidence ≥ 90.00% and absolute Δ mean >= ±8.87%: Fine details of change detection per experiment.
|
Closes #12859
HttpErrors for datadog_logs sink are now retry-able.Note: Now that the two sinks have similar error reporting behavior, there is a little bit more repeated code. I stopped from more refactoring to pull out common code into the
sinks/datadog/mod.rsas I wasn't sure how far we wanted to take it... For example the RetryLogic implementations are almost the same. And the Service'scall()function have some commonalities. I'm generally a fan of reducing as much redundant code as possible but I realize it may not be desirable in all scenarios.