This reduces scrape duration for targets returning big responses.
The response body was already read into memory in stream parsing mode before this change,
so this commit shouldn't increase memory usage.
- Rename logDebug() to logDebugf() and pass format string together
with format args directly to logDebugf(). This eliminates fmt.Sprintf()
overhead at logDebug() call site when debugging is disabled.
- Format labels in debug message in Prometheus format, e.g. {label1="value1",...labelN="valueN"}
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3025
* vmalert: add `debug` mode for alerting rules
Debug information includes alerts state changes and requests
sent to the datasource. Debug can be enabled only on rule's
level. It might be useful for debugging unexpected
behaviour of alerting rule.
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3025
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* vmalert: review fixes
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* Update app/vmalert/alerting.go
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* vmalert: go fmt
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
- Add getCommonParamsWithDefaultDuration function and use it at /api/v1/series, /api/v1/labels and /api/v1/label/.../values
- Document the default behaviour for setting 5 minutes time range if start arg isn't passed to /api/v1/series, /api/v1/labels and /api/v1/label/.../values
- Document the change at docs/CHANGELOG.md
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/pull/3052
The panic may trigger during data blocks' processing received
from vmstorage nodes when some of vmstorage nodes return an error
or when `-replicationFactor` is set to values higher than 2 at `vmselect`.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3058
Note that the parallel execution of `union()` args may take more memory and CPU time
than the sequential execution if args contain heavy queries, which may load all the available CPU,
disk and memory resources and vmselect and vmstorage levels.
- Use getScalar() function for obtaining the expected scalar from phi arg
- Reduce the error message returned to the user when incorrect phi is passed to histogram_quantiles
- Improve the description of this bugfix in the docs/CHANGELOG.md
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3026
Cache sanitized label names and return them next time.
This reduces the number of allocations and speeds up the SanitizeLabelName()
function for common case when the number of unique label names is smaller than 100k
The following regex patterns are optimized:
- literal string match, e.g. "foo"
- prefix match, e.g. "foo.*" and "foo.+"
- substring match, e.g. ".*foo.*" and ".+foo.+"
- alternate values match, e.g. "foo|bar|baz"
- Document the change at docs/CHANGELOG.md
- Move auth token parsing from app/vmagent/opentsdbhttp/ to app/vmagent/main.go,
since it must be parsed only when multitenancy support is enabled at vmagent side.
See https://docs.victoriametrics.com/vmagent.html#multitenancy
These metrics allow alerting when the number of unique series approach the limit.
For example, the following query alerts when the number of series reaches 90% of the configured limit:
vm_hourly_series_limit_current_series / vm_hourly_series_limit_max_series > 0.9
We switch default alert's source link to redirect user
to vmalert's UI instead of previous JSON object. While it breaks
compatibility, it also supposed to improve user's experience.
The old behavior can be achieved by updating `-external.alert.source`
command-line flag.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Signed-off-by: hagen1778 <roman@victoriametrics.com>
The 429 status code means that the server is overwhelmed with requests.
The client can retry the request after some wait time.
Implement this strategy for service discovery and scrape requests.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2940
* Explicitly store a pointer to UserReadableError in the error interface.
Previously Go automatically converted the value to a pointer before storing in the error interface.
* Add Unwrap() method to UserReadableError, so it can be used transparently with the other code,
which calls errors.Is() and errors.As().
* Document the change in docs/CHANGELOG.md
Previously a single syncwg.WaitGroup was used for tracking the lifetime of processBlock callbacks
across all the per-vmstorage goroutines. This could be slow on systems with many CPU cores
because of inter-CPU synchronization overhead.
Use a separate per-vmstorage sync.WaitGroup instead in order to reduce inter-CPU synchronization overhead.
This should imrpove performance for heavy queries over big number of blocks on multi-CPU systems.
* lib/storage: prevent excessive loops when storage is in RO
Returning nil error when storage is in RO mode results
into excessive loops and function calls which could
result into CPU exhaustion. Returning an err instead
will trigger delays in the for loop and save some resources.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* document the change
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
The `-mod=vendor` is automatically set when there is a `vendor` directory
starting from Go1.14 - see https://go.dev/doc/go1.14#go-command
Since the minimum supported Go version for VictoriaMetrics is Go1.17,
then the `-mod=vendor` option is no longer needed.
Reduce inter-CPU communications when processing the query over big number of time series.
This should improve performance for queries over big number of time series
on systems with many CPU cores.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2896
Based on b596ac3745
Thanks to @zqyzyq for the idea.
Usernames could be duplicate if it has uniq password.
vmauth makes routing based on auth token and username + password combination must be unique for this case.
- Use binary search instead of linear scan when locating the run of smallest timestamps
in blocks with intersected time ranges. This should improve performance
when merging blocks with big number of samples
- Skip samples with duplicate timestamps. This should increase query performance
in cluster version of VictoriaMetrics with the enabled replication.
* vmalert: deprecate alert's status link
Deprecate alert's status link `/api/v1/<groupID>/<alertID>/status` in favour of
`api/v1/alerts?group_id=<group_id>&alert_id=<alert_id>"`.
The change was needed for simplifying logic in vmselect for proxying vmalert's requests.
The old alert's status link will be still supported for a few versions but will be removed in the future.
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2825
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* vmalert: fix review comments
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Previously the time series could be put into dateMetricIDCache without
registering in the per-day inverted index if GetOrCreateTSIDByName
finds TSID entry in the global index. This could lead to missing
series in query results.
The issue has been introduced in the commit 55e7afae3a,
which has been included in VictoriaMetrics v1.78.0
* docs: warn about potential issue with read queries for 1.78.0
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* docs: warn about potential issue with read queries for 1.78.0
Signed-off-by: hagen1778 <roman@victoriametrics.com>
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 the cache could store 10K unique regexps. When every regexp is huge (e.g. hundreds of kilobytes),
then the total cache size could grow to multiples of gigabytes. Now the cache size is limited by the total length
of all cached regexps. So huge regexps won't result in high memory usage for the cache.
If the previous dial attempt was unsuccessful, then all the new dial attempts are skipped
until the background goroutine determines that the given address can be successfully dialed.
This reduces query latency when some of vmstorage nodes are unavailable and dialing them is slow.
This should help with https://github.com/VictoriaMetrics/VictoriaMetrics/issues/711
This commit is based on ideas from the https://github.com/VictoriaMetrics/VictoriaMetrics/pull/2756
The main differences are:
- The check for healthy/unhealthy storage nodes is moved one level lower from app/vmselect/netstorage to lib/netutil.ConnPool.
This makes possible re-using this feature everywhere lib/netutil.ConnPool is used.
- The check doesn't take into account handshake errors for already established connections.
Handshake errors usually mean improperly configured VictoriaMetrics cluster, so they shouldn't be ignored.
* vmselect: limit `end` param max value by 2d in future
The change is applied only to service handlers like `/labels` or `/series`
and limits the `end` param by max value <= now() + 2 days. The same limit
is applied for the ingested data, so no reason to allow to request data
in future far than that.
The change is also needed for corner cases like https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2669
where too high `end` value triggers inefficient global index search.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* docs/CHANGELOG.md: document the bugfix
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* feat: make datepicker to be set to last 30 min by default
* fix: correct spinner while loading data
* feat: change legend style
* app/vmselect: `make vmui-update`
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
- Remove unused js bloatware from /targets page. This strips down binary size by more than 100Kb
- Add /service-discovery page for API compatibility with Prometheus
- Properly load bootstrap.min.css from /prometheus/targets
- Serve static contents for /targets page from app/vminsert instead of app/vmselect, because /targets page is served from there
The default size of `indexdb/tagFilters` now can be overridden via
`storage.cacheSizeIndexDBTagFilters` flag.
Please, be careful with changing default size since it may
lead to inefficient work of the vmstorage or OOM exceptions.
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2663
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Co-authored-by: Nikolay <nik@victoriametrics.com>
* promrelabel: add support of `lowercase` and `uppercase` relabeling actions
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2664
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* lib/storage: make golangci-lint happy
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Co-authored-by: Nikolay <nik@victoriametrics.com>
* lib/promscrape/discovery/kubernetes: properly updates discovered scrape works
previously, added or updated scrapeworks may override previuosly
discovered.
it happens because swosByKey may contain small subset of kubernetes
objects with it's labels.
It happens for objectsUpdated and objectsAdded maps, which include only changed elements
* Properly calculate vm_promscrape_discovery_kubernetes_scrape_works
Co-authored-by: f41gh7 <nik@victoriametrics.com>
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
Unexpectedly, Grafana makes an extra request to `/rules`
handler in addition to `/api/v1/rules` calls in alerts UI.
This happens only for Grafana versions older than 8.5.*.
Apparently, this is related to support of other monitoring
systems.
Prometheus responds with `text/html` content for UI page `/rules`
to such requests. Actually, returning just a blank page with
SC=200 works as well.
Returning actual response of `/api/v1/rules`
results in error in Grafana since it expects a `yaml` (?) in response.
So we add a placeholder to `vmalert`.
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2583
Signed-off-by: hagen1778 <roman@victoriametrics.com>
This should reduce potential spikes in the number of established connections in the following cases:
- when the connection establishing procedure becomes temporarily slow
- after a temporary spike in the rate of ConnPool.Get() calls
See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2552
Previously ScrapeConfig.clone() was improperly copying promauth.Secret fields -
their contents was replaced with `<secret>` value.
This led to inability to use passwords and secrets in `-promscrape.config` file.
The bug has been introduced in v1.77.0 in the commit 67b10896d2
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2551
* {lib/promscrape,app/vmagent}: adds sigv4 support for vmagent remoteWrite
moves aws related code into separate lib from lib/promscrape
it allows to write data from vmagent to the AWS managed prometheus (cortex)
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1287
* Apply suggestions from code review
* wip
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* lib/{storage,flagutil} - Add option for snapshot autoremoval
- add prometheus-like duration as command flag
- add option to delete stale snapshots
- update duration.go flag to re-use own code
* wip
* lib/flagutil: re-use Duration.Set() call in NewDuration
* wip
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* Export "null" in jsonl instead of NaN
The NaN appeared because of staleness markers that were added for compatibility. I think it's better to use json `null`, implemented here.
Also maybe it also makes sense to add a flag like `?skip-staleness-markers=true` to `/export`, to skip nulls at all?
* Update app/vmselect/prometheus/export.qtpl
* app/vmselect/prometheus/export.qtpl.go: `make quicktemplate-gen`
* docs/CHANGELOG.md: document the change
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
This should prevent from `duplicate time series` errors when executing the following query:
kube_pod_container_resource_requests{resource="cpu"} * on (namespace,pod) group_left() (kube_pod_status_phase{phase=~"Pending|Running"}==1)
where `kube_pod_status_phase{phase=~"Pending|Running"}==1` filters out diplicate time series
Reduce the number of memory allocations in this function. This improves its performance by up to 50%.
This should improve service discovery speed when big number of potential targets with big number of meta-labels
are generated by service discovery.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2270
* lib/promscrape: adds job restart method
it must restart only ScrapeConfig with changed content
this change greatly reduce time, that needed for job restart
and it should decrease possible data loss when config frequently changed at kubernetes based deployments
Apply suggestions from code review
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
* wip
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
The following actions are taken:
- Increase the TLS hashdshake timeout from 5 seconds to 10 seconds
- Increase dial timeout from 5 seconds to 30 seconds
- Specify DialContext instead of Dial in http.Transport. This allows properly handling
the Context arg during dialing the remote storage
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1699
Some of the releases could negatively affect performance for a limited
period of time due to some changes in core. Update details are meant to
warn users about expected changes in peformance after the update.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* fixes jwt token parse with correct base64Url decoding
it must be applied according to jwt RFC that requires token to be URL safe
added slow path for decoding tokens with std base64 decoding
adds error logging for vmgateway
* docs/CHANGELOG.md: document the bugfix
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* lib/discovery/consul: update services on the watcher's start
Previously, watcher's start was only initing goroutines for discovery
but not waiting for the first iteration to end. It means first Consul
discovery wasn't returning discovered targets until the next iteration.
The change makes the watcher's start blocking until we get first discovery
iteration done and all registries updated.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* vmalert: remove workarounds for consul SD
Now when consul SD lib properly updates services
on the first start, we don't need workarounds in vmalert.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* lib/discovery/consul: update after review
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* wip
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* lib/index: reduce read/write load after indexDB rotation
IndexDB in VM is responsible for storing TSID - ID's used for identifying
time series. The index is stored on disk and used by both ingestion and read path.
IndexDB is stored separately to data parts and is global for all stored data.
It can't be deleted partially as VM deletes data parts. Instead, indexDB is
rotated once in `retention` interval.
The rotation procedure means that `current` indexDB becomes `previous`,
and new freshly created indexDB struct becomes `current`. So in any time,
VM holds indexDB for current and previous retention periods.
When time series is ingested or queried, VM checks if its TSID is present
in `current` indexDB. If it is missing, it checks the `previous` indexDB.
If TSID was found, it gets copied to the `current` indexDB. In this way
`current` indexDB stores only series which were active during the retention
period.
To improve indexDB lookups, VM uses a cache layer called `tsidCache`. Both
write and read path consult `tsidCache` and on miss the relad lookup happens.
When rotation happens, VM resets the `tsidCache`. This is needed for ingestion
path to trigger `current` indexDB re-population. Since index re-population
requires additional resources, every index rotation event may cause some extra
load on CPU and disk. While it may be unnoticeable for most of the cases,
for systems with very high number of unique series each rotation may lead
to performance degradation for some period of time.
This PR makes an attempt to smooth out resource usage after the rotation.
The changes are following:
1. `tsidCache` is no longer reset after the rotation;
2. Instead, each entry in `tsidCache` gains a notion of indexDB to which
they belong;
3. On ingestion path after the rotation we check if requested TSID was
found in `tsidCache`. Then we have 3 branches:
3.1 Fast path. It was found, and belongs to the `current` indexDB. Return TSID.
3.2 Slow path. It wasn't found, so we generate it from scratch,
add to `current` indexDB, add it to `tsidCache`.
3.3 Smooth path. It was found but does not belong to the `current` indexDB.
In this case, we add it to the `current` indexDB with some probability.
The probability is based on time passed since the last rotation with some threshold.
The more time has passed since rotation the higher is chance to re-populate `current` indexDB.
The default re-population interval in this PR is set to `1h`, during which entries from
`previous` index supposed to slowly re-populate `current` index.
The new metric `vm_timeseries_repopulated_total` was added to identify how many TSIDs
were moved from `previous` indexDB to the `current` indexDB. This metric supposed to
grow only during the first `1h` after the last rotation.
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1401
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* wip
* wip
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* lib/promscrape: support prometheus-like duration in scrape configs
The change allows to specify duration values like `1d`, `1w`
for fields `scrape_interval`, `scrape_timeout`, etc.
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/817#issuecomment-1033384766
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* lib/blockcache: make linter happy
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* lib/promscrape: support prometheus-like duration in scrape configs
* add support for extra fields `scrape_align_interval` and `scrape_offset`;
* support Prometheus duration parsing for `__scrape_interval__`
and `__scrape_duration__` labels;
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* wip
* wip
* docs/CHANGELOG.md: document the feature
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* fixes service discovery for kubernetes
now it must take in account all pods that belong to the discovered endpoint and endpointslice
adds simple test for endpoints
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2134
* wip
* docs/CHANGELOG.md: document the change
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* adds CGO build for arm64
it must improve performance for arm64 based deployments of vmstorage and
vmsingle for 15-20%
it depends on gozstd package update for correct musl gozstd vendoring
* typo fixes
* docs/CHANGELOG.md: document the change
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* Mention about the ability to configure vmalert notifiers via files in docs/CHANGELOG.md
* Mention about the ability to use Consul service discovery for vmalert notifiers in docs/CHANGELOG.md
* Run `make docs-sync` in order to sync app/vmalert/README.md to docs/vmalert.md
This metric shows the number of CPU cores available to the process.
This allows creating alerting rules on CPU saturation with the following query:
rate(process_cpu_seconds_total[5m]) / process_cpu_cores_available > 0.9
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2107
This adds more optimization cases for https://utcc.utoronto.ca/~cks/space/blog/sysadmin/PrometheusLabelNonOptimization
For example:
* Multi-level transform functions. For example, abs(round(foo{a="b"})) + bar{x="y"}
is now optimized to abs(round(foo{a="b",x="y"})) + bar{a="b",x="y"}
* Binary operations with `on()`, `without()`, `group_left()` and `group_right()` modifiers.
For example, foo{a="b"} on (a) + bar is now optimized to foo{a="b"} on (a) + bar{a="b"}
* Multi-level binary operations. For example, foo{a="b"} + bar{x="y"} + baz{z="q"}
is now optimized to foo{a="b",x="y",z="q"} + bar{a="b",x="y",z="q"} + baz{a="b",x="y",z="q"}
* Aggregate functions. For example, sum(foo{a="b"}) by (c) + bar{c="d"}
is now optimized to sum(foo{a="b",c="d"}) by (c) + bar{c="d"}
* fix: add date validate for time range
* app/vmselect/vmui: `make vmui-update`
* docs/CHANGELOG.md: document the bugfix
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
Previously these caches could exceed limits set via `-memory.allowedPercent` and/or `-memory.allowedBytes`,
since limits were set independently per each data part. If the number of data parts was big, then limits could be exceeded,
which could result to out of memory errors.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2007
