It allows to modify exist series values.
User must write modified series into vminsert API
/insert/0/prometheus/api/v1/update/series
vminsert will generate id and add it to the series as __generation_id
label.
Modified series merged at vmselect side.
Only last series modify request at given time range will be applied.
Modification request could be exported with the following API request:
`curl localhost:8481/select/0/prometheus/api/v1/export -g -d
'reduce_mem_usage=true' -d 'match[]={__generation_id!=""}'`
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/844
adds guide
allow single datapoint modification
vmselectapi: prevent MetricBlockRef corruption
Modofying of MetricName byte slice may result into MetricBlockRef
corruption, since `ctx.mb.MetricName` is a pointer to
`MetricBlockRef.MetricName`.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Revert "vmselectapi: prevent MetricBlockRef corruption"
This reverts commit cf36bfa1895885fcc7dc2673248ee56c78180ea0.
app/vmstorage/servers: properly copy MetricName into MetricBlock inside blockIterator.NextBlock
This should fix the issue at cf36bfa189
(cherry picked from commit 916f1ab86c)
app/vmselect: correctly update single datapoint at merge
app/vmselect: adds mutex for series update map
previously it was sync api, but function signature was changed for performance optimizations
Previously SearchMetricNames was returning unmarshaled metric names.
This wasn't great for vmstorage, which should spend additional CPU time
for marshaling the metric names before sending them to vmselect.
While at it, remove possible duplicate metric names, which could occur when
multiple samples for new time series are ingested via concurrent requests.
Also sort the metric names before returning them to the client.
This simplifies debugging of the returned metric names across repeated requests to /api/v1/series
Previously bytesutil.Resize() was copying the original byte slice contents to a newly allocated slice.
This wasted CPU cycles and memory bandwidth in some places, where the original slice contents wasn't needed
after slize resizing. Switch such places to bytesutil.ResizeNoCopy().
Rename the original bytesutil.Resize() function to bytesutil.ResizeWithCopy() for the sake of improved readability.
Additionally, allocate new slice with `make()` instead of `append()`. This guarantees that the capacity of the allocated slice
exactly matches the requested size. The `append()` could return a slice with bigger capacity as an optimization for further `append()` calls.
This could result in excess memory usage when the returned byte slice was cached (for instance, in lib/blockcache).
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2007
This should improve lookup performance if the same `label=value` pair exists
in big number of time series.
This should also reduce memory usage for mergeset data cache, since `tag->metricIDs` rows
occupy less space than the original `tag->metricID` rows.
