Troubleshooting #

This document contains troubleshooting guides for the most common issues when working with VictoriaMetrics.

General troubleshooting checklist #

If you encounter an issue or have a question about VictoriaMetrics components, follow these steps to quickly find a solution:

1. Check the version of the VictoriaMetrics component you are troubleshooting and compare it with the latest available version .

   If you are running an older version, the issue may already be fixed. Review the changelog for all releases between your version and the latest release to see whether the problem has been resolved.

   If the issue is fixed in a newer release, upgrade and verify that the problem no longer occurs:

   • How to upgrade single-node VictoriaMetrics
   • How to upgrade VictoriaMetrics cluster

   The upgrade procedure for other VictoriaMetrics components is as simple as gracefully stopping the component by sending it a SIGINT signal and starting the new version of the component.

   In rare cases, upgrades may include breaking changes. These cases are documented in the changelog , especially check the Update notes near the top of the changelog, as they point out any special actions or considerations to take when upgrading.

2. Review command-line flags passed to VictoriaMetrics components and remove any flags whose impact on your workload is unclear. VictoriaMetrics components are optimized to work well with default settings (that is, when flags aren’t explicitly set). Unnecessary or poorly understood flags can lead to unexpected behavior, so it’s best to remove them unless you clearly understand why they are needed.

3. Check logs. They often contain useful details about the root cause and possible fixes.

   If the logs don’t provide enough information, try searching the error message on Google. In many cases, the issue has already been discussed (in documentation, on Stack Overflow, or in GitHub issues), and a solution may already be available.

4. If VictoriaMetrics logs do not have relevant information, then try searching for the issue on Google using multiple keywords and phrases specific to the issue. In many cases, both the issue and its solution are already documented.

5. Try searching for the issue in VictoriaMetrics GitHub . The signal-to-noise ratio of search results here is much lower than on Google, but sometimes it can help find relevant information when Google fails. If you located the relevant GitHub issue, but it lacks details for diagnosis or troubleshooting, then please add them in the issue comments. This increases the chance that it will be resolved soon.

6. Try searching for information about the issue in the VictoriaMetrics source code . GitHub code search may not be very effective in some cases, so it is recommended to check out the VictoriaMetrics source code and perform a local search in the code. Note that the source code for the VictoriaMetrics cluster is located in the cluster branch.

7. If the steps above didn’t help to find the solution to the issue, then please file a new issue with as many details as possible on how to reproduce it.

   After that you can post the link to the issue in the VictoriaMetrics Slack chat , so the VictoriaMetrics community can help find a solution. It is better to file the issue on VictoriaMetrics GitHub before posting your question to the VictoriaMetrics Slack chat, since GitHub issues are indexed by Google, while Slack messages are not. This simplifies finding a solution to the issue for future VictoriaMetrics users.

8. Pro tip 1: if you see that VictoriaMetrics docs contain incomplete or incorrect information, then please create a pull request with the relevant changes or a new issue explaining the problem. This will help the VictoriaMetrics community.

9. Pro tip 2: please provide links to existing docs / GitHub issues / StackOverflow questions instead of copying and pasting the information from these sources when asking or answering questions to the VictoriaMetrics community. If the linked resources do not have enough information, then it is better to add the missing information to the original web resource before linking it to Slack chat. This will simplify searching for this information in the future for VictoriaMetrics users via Google and Perplexity .

10. Pro tip 3: if you are answering somebody’s question about VictoriaMetrics components in GitHub issues / Slack chat / StackOverflow, then the best answer is a direct link to the information with the answer or solution to the question. The better answer is a concise message with multiple links to the relevant information. The worst answer is a message with misleading or completely wrong information.

11. Pro tip 4: If you can fix the issue on your own, then please do it and provide the corresponding pull request! We are happy to get pull requests from the VictoriaMetrics community.

Unexpected query results #

If you see unexpected or unreliable query results from VictoriaMetrics, then try the following steps:

1. Check whether simplified queries return unexpected results. For example, if the query looks like sum(rate(http_requests_total[5m])) by (job), then check whether the following queries return expected results:

   • Remove the outer sum and execute rate(http_requests_total[5m]). Aggregations could hide missing series, data gaps, or anomalies.

   • If the query returns too many series, try adding more specific label filters. For example, if you see that the original query returns unexpected results for the job="foo", then use the rate(http_requests_total{job="foo"}[5m]) query. Continue adding more specific label filters until the resulting query returns a manageable number of time series.

   • Remove the outer rate and execute http_requests_total. Add label filters to reduce the number of returned series if needed.

   Sometimes the query may be improperly constructed, leading to unexpected results. It is recommended to read and understand MetricsQL docs , especially subqueries and rollup functions sections.

2. If the simplest query continues returning unexpected / unreliable results, then try verifying correctness of raw unprocessed samples in vmui via the Raw Query tab.

   Responses returned from /api/v1/query and /api/v1/query_range contain evaluated data instead of stored raw samples. In some cases, staleness , deduplication , or irregular scrapes can affect evaluations. See this short video for details.

   Raw data can be downloaded via the Export button in vmui’s Raw Query tab or via /api/v1/export query on the given [start..end] time range and check whether they are expected:

         single-node: curl http://victoriametrics:8428/api/v1/export -d 'match[]=http_requests_total' -d 'start=...' -d 'end=...' -d 'reduce_mem_usage=1'
   
   cluster: curl http://<vmselect>:8481/select/<tenantID>/prometheus/api/v1/export -d 'match[]=http_requests_total' -d 'start=...' -d 'end=...' -d 'reduce_mem_usage=1'
       

   When raising a GitHub ticket about query issues, please also attach the raw data, so maintainers can reproduce your case locally.

3. Try executing the query with tracer enabled. The trace contains a lot of additional information about query execution, series matching, caches, and internal modifications. When raising a GitHub ticket about query issues, please also attach the trace so maintainers can investigate.

4. If you observe gaps when plotting series, it is likely caused by irregular intervals for metrics collection (network delays or targets unavailability during scrapes, irregular pushes, irregular timestamps). VictoriaMetrics automatically fills the gaps based on the median interval between data samples . This may yield incorrect results for irregular data, as the median will be skewed. In this case, it is recommended to either fix the irregularities or switch to the static interval for gaps filling by setting -search.minStalenessInterval=5m command-line flag (5m is used by Prometheus by default).

5. Sometimes, response caching may lead to unexpected results when samples with older timestamps are ingested into VictoriaMetrics (aka backfilling ). Try disabling response cache and see whether this helps:

   • By clicking on the toggle Disable cache in vmui.

   • By passing -search.disableCache command-line flag to a single-node VictoriaMetrics or to all the vmselect components if the cluster version of VictoriaMetrics is used.

   • By passing nocache=1 query arg to every request to /api/v1/query and /api/v1/query_range. If you use Grafana, then this query arg can be specified in the Custom Query Parameters field in Prometheus datasource settings. See these docs for details.

   If the problem was in the cache, try resetting it via the resetRollupCache handler .

6. Cluster version of VictoriaMetrics may return partial responses by default when some of the vmstorage nodes are temporarily unavailable. See cluster availability docs . If you want to prioritize query consistency over cluster availability, then pass -search.denyPartialResponse command-line flag to all the vmselect nodes. This causes VictoriaMetrics to return an error during query execution if at least one vmstorage node is unavailable. Another option is to pass deny_partial_response=1 query argument to /api/v1/query and /api/v1/query_range. If you use Grafana, then this query argument can be specified in the Custom Query Parameters field in Prometheus/VictoriaMetrics datasource settings. See these docs for details.

7. If you pass the -replicationFactor command-line flag to vmselect, then it is recommended to remove this flag from vmselect, since it may lead to incomplete responses when vmstorage nodes contain less than -replicationFactor copies of the requested data.

8. If you observe that recently written data is not immediately visible/queryable, then read more about query latency behavior.

9. Try upgrading to the latest available version of VictoriaMetrics and verifying whether the issue is fixed there.

10. Inspect command-line flags passed to VictoriaMetrics components. If you don’t clearly understand the purpose or the effect of some flags, then remove them from the list of flags. VictoriaMetrics components are optimized to work well with default settings (that is, when flags aren’t explicitly set). Unnecessary or poorly understood flags can lead to unexpected behavior, so it’s best to remove them unless you clearly understand why they are needed.

11. If the steps above didn’t help identify the root cause of unexpected query results, then file a bug report with details on how to reproduce the issue. Instead of sharing screenshots in the issue, consider sharing the query, raw samples and trace results via VMUI .

Slow data ingestion #

These are the most common reasons for slow data ingestion in VictoriaMetrics:

1. Memory shortage for the given amounts of active time series .

   VictoriaMetrics (or vmstorage in the cluster version of VictoriaMetrics) maintains an in-memory cache storage/tsid for a quick search for internal series IDs for each incoming metric. VictoriaMetrics automatically determines the maximum size for this cache depending on the available memory on the host where VictoriaMetrics (or vmstorage) runs. If the cache size isn’t enough to hold all the entries for active time series, then VictoriaMetrics locates the required data on disk, unpacks it, reconstructs the missing entry, and adds it to the cache. This takes additional CPU time and disk read I/O.

   The official Grafana dashboards for VictoriaMetrics contain a Slow inserts graph that shows the cache miss percentage for the storage/tsid cache during data ingestion. If the slow inserts graph shows values greater than 5% for more than 10 minutes, then it is likely that the current number of active time series cannot fit the storage/tsid cache.

   These are the solutions that exist for this issue:

   • Increase the available memory on the host where VictoriaMetrics runs until the slow inserts percentage drops to 5% or less. If you run a VictoriaMetrics cluster, then you need to increase the total available memory at all vmstorage nodes. This can be done in two ways: either to increase the available memory for each vmstorage node or to add more vmstorage nodes to the cluster to spread the load.

   • Reduce the number of active time series. The official Grafana dashboards for VictoriaMetrics contain a graph showing the number of active time series. Use the cardinality explorer to determine and fix the source of high cardinality .

   • Insert performance can degrade when the same time series arrives with labels in a different order. Ensure your ingestion client always sends labels in a consistent order for each series. Prometheus and vmagent already guarantee this, but custom or third-party clients might not. As a fallback, you can enable -sortLabels=true on VictoriaMetrics or on vminsert in cluster mode. This forces the server to normalize label order, though it increases CPU usage during ingestion.

2. High churn rate . When VictoriaMetrics encounters a sample for a new time series, it needs to register the time series in the internal index (aka indexdb), so it can be quickly located during select queries. The process of registering new time series in the internal index is an order of magnitude slower than the process of adding a new sample to an already registered time series. So VictoriaMetrics may work slower than expected under high churn rate .

   The official Grafana dashboards for VictoriaMetrics provide a Churn rate graph, which shows the average number of new time series registered during the last 24 hours. If this number exceeds the number of active time series , then you need to identify and fix the source of high churn rate . The most common source of high churn rate is a label that frequently changes its value (like timestamp, session_id). Try avoiding such labels. The cardinality explorer can help identify such labels.

3. Resource shortage. The official Grafana dashboards for VictoriaMetrics contain Resource usage graphs that show memory usage, CPU usage, disk I/O usage, etc. Make sure VictoriaMetrics has enough free resources for gracefully handling potential spikes in workload according to the following recommendations:

   • 50% of free CPU
   • 50% of free memory
   • 20% of free disk space

   If VictoriaMetrics components have lower amounts of free resources, then this may lead to significant performance degradation when workload increases slightly. For example:

   • If the percentage of free CPU is close to 0, then VictoriaMetrics may experience arbitrarily long delays during data ingestion, even with slight increases in ingestion rate.

   • If the percentage of free memory reaches 0, then the Operating System where VictoriaMetrics components run may not have enough memory for the page cache . VictoriaMetrics relies on the page cache for quick queries over recently ingested data. If the operating system does not have enough free memory for the page cache, then it must re-read the requested data from disk. This may significantly increase disk read I/O and slow down both queries and data ingestion.

   • If free disk space is below 20%, then VictoriaMetrics may be unable to perform optimal background merge of the incoming data. This results in more data files on disk. That, in turn, slows down both data ingestion and querying. See these docs for details.

4. If you run the cluster version of VictoriaMetrics, then make sure vminsert and vmstorage components are located in the same network with a low network latency between them. vminsert packs incoming data into batch packets and sends them to vmstorage one by one. It waits until vmstorage returns back an ack response before sending the next packet. If the network latency between vminsert and vmstorage is high (for example, if they run in different datacenters), then this may become a limiting factor for data ingestion speed.

   The official Grafana dashboard for the cluster version of VictoriaMetrics contains a connection saturation panel for vminsert components. If this graph reaches 100% (1s), then it is likely you have issues with network latency between vminsert and vmstorage. Another possible issue for 100% connection saturation between vminsert and vmstorage is a resource shortage in the vmstorage nodes. In this case, you need to increase the amount of available resources (CPU, RAM, disk I/O) at vmstorage nodes or add more vmstorage nodes to the cluster.

5. Noisy neighbor. Make sure VictoriaMetrics components run in an environment without other resource-hungry apps. Such apps may steal RAM, CPU, disk I/O, and network bandwidth that are needed for VictoriaMetrics components. Issues like this are hard to catch via the official Grafana dashboard for the cluster version of VictoriaMetrics and proper diagnosis would require checking resource usage on the instances where VictoriaMetrics runs.

6. If you see a TooHighSlowInsertsRate alert when single-node VictoriaMetrics or vmstorage has enough free CPU and RAM, then increase the -cacheExpireDuration command-line flag at single-node VictoriaMetrics or at vmstorage to a value that exceeds the interval between ingested samples for the same time series (aka scrape_interval). See this comment for more details.

7. If you see constant and abnormally high CPU usage for the VictoriaMetrics component, check the CPU spent on GC panel on the corresponding Grafana dashboard in the Resource usage section. If the percentage of CPU time spent on garbage collection is high, then CPU usage of the component can be reduced at the cost of higher memory usage by increasing the GOGC environment variable. By default, VictoriaMetrics components use GOGC=30. Try running VictoriaMetrics components with GOGC=100 and see whether this helps reduce CPU usage. Note that higher GOGC values may increase memory usage.

Slow queries #

Some queries may take more time and resources (CPU, RAM, network bandwidth) than others. VictoriaMetrics logs slow queries if their execution time exceeds the duration passed to -search.logSlowQueryDuration command-line flag (5s by default).

VictoriaMetrics provides a top queries page in VMUI that shows the longest-running queries. And Query execution stats for dumping slow queries to logs.

These are the solutions that exist for improving the performance of slow queries:

• Investigating the bottleneck in query execution using query tracing . It will show the percentage of time spent on each execution step and help understand the volume of processed data.

• Adding more CPU and memory to VictoriaMetrics, so it may perform the slow query faster. If you use the cluster version of VictoriaMetrics, then migrating vmselect nodes to machines with more CPU and RAM should help improve speed for slow queries. Query performance is always limited by the resources of one vmselect that processes the query. For example, if 2 vCPU cores on vmselect can’t process queries fast enough, then migrating vmselect to a machine with 4 vCPU cores should increase heavy query performance by up to 2x. If the line on the concurrent select graph from the official Grafana dashboard for VictoriaMetrics is close to the limit, then prefer adding more vmselect nodes to the cluster. Sometimes adding more vmstorage nodes can also help improve the speed for slow queries.

• Rewriting slow queries, so they become faster.

  The main source of slow queries in practice is alerting and recording rules with long lookbehind windows in square brackets. These queries are frequently used in SLI/SLO calculations such as Sloth .

  For example, avg_over_time(up[30d]) > 0.99 needs to read and process all the raw samples for the up time series over the last 30 days each time it executes. If this query is executed frequently, it can take a significant share of CPU, disk read I/O, network bandwidth, and RAM. Such queries can be optimized in the following ways:

  • To reduce the look-behind window in square brackets. For example, avg_over_time(up[10d]) takes up to 3x less compute resources than avg_over_time(up[30d]) at VictoriaMetrics.
  • To increase the evaluation interval for alerting and recording rules, so they are executed less frequently. For example, increasing the -evaluationInterval command-line flag value at vmalert from 1m to 2m should reduce compute resource usage by VictoriaMetrics 2x.

  Another source of slow queries is improper use of subqueries . It is recommended to avoid subqueries if you don’t clearly understand how they work. It is easy to create a subquery without knowing about it. For example, rate(sum(some_metric)) is implicitly transformed into the following subquery according to implicit conversion rules for MetricsQL queries :

        rate(
    sum(
      default_rollup(some_metric[1i])
    )[1i:1i]
  )
      

  It is likely this query won’t return the expected results. Instead, sum(rate(some_metric)) must be used instead. See this article for more details.

  See also this article , which explains how to identify and optimize slow queries.

Out of memory errors #

The following are the most common sources of out-of-memory (aka OOM) crashes in VictoriaMetrics:

1. Improper command-line flag values. Inspect command-line flags passed to VictoriaMetrics components. If you don’t clearly understand the purpose or the effect of some flags, remove them from the list of flags passed to VictoriaMetrics components. Improper command-line flag values may lead to increased memory and CPU usage. Increased memory usage increases the risk of OOM crashes. VictoriaMetrics is optimized to run with default flag values (e.g., when they aren’t explicitly set).

   For example, it isn’t recommended to change cache sizes in VictoriaMetrics, as this frequently leads to OOM exceptions. These docs refer to command-line flags that aren’t recommended to tune. If you see that VictoriaMetrics needs to increase some cache sizes for the current workload, then it is better to migrate to a host with more memory instead of trying to tune cache sizes manually.

2. Unexpected heavy queries. The query is considered heavy if it needs to select and process millions of unique time series. Such a query may cause an OOM exception, as VictoriaMetrics needs to keep some per-series data in memory. VictoriaMetrics provides various settings that can help limit resource usage. For more context, see How to optimize PromQL and MetricsQL queries . VictoriaMetrics also provides query tracer to help identify the source of heavy queries. Slow queries can be logged with additional details via Query execution stats .

3. Lack of free memory for processing workload spikes. If VictoriaMetrics components use almost all the available memory under the current workload, then it is recommended to migrate to a host with larger amounts of memory. This would protect from possible OOM crashes on workload spikes. It is recommended to have at least 50% of free memory to gracefully handle possible workload spikes. See capacity planning for single-node VictoriaMetrics and capacity planning for the cluster version of VictoriaMetrics .

Cluster instability #

The VictoriaMetrics cluster may become unstable if there are not enough free resources (CPU, RAM, disk I/O, network bandwidth) for processing the current workload.

The most common sources of cluster instability are:

• Workload spikes. For example, if the number of active time series increases by 2x while the cluster does not have enough free resources for processing the increased workload, then it may become unstable. VictoriaMetrics provides several configuration settings to limit unexpected workload spikes. See these docs for details.

• Various maintenance tasks, such as rolling upgrades or rolling restarts, during configuration changes. For example, if a cluster contains N=3 vmstorage nodes and they are restarted one-by-one (aka rolling restart), then the cluster will have only N-1=2 healthy vmstorage nodes during the rolling restart. This means that the load on healthy vmstorage nodes increases by at least 100%/(N-1)=50% compared to the load before rolling restart. E.g., they need to process 50% more incoming data and to return 50% more data during queries. In reality, the load on the remaining vmstorage nodes increases even more because they need to register new time series that were re-routed from a temporarily unavailable vmstorage node. If vmstorage nodes had less than 50% of free resources (CPU, RAM, disk I/O) before the rolling restart, then it can lead to cluster overload and instability for both data ingestion and querying.

  The workload increase during rolling restart can be reduced by increasing the number of vmstorage nodes in the cluster. For example, if the VictoriaMetrics cluster contains N=11 vmstorage nodes, then the workload increase during rolling restart of vmstorage nodes would be 100%/(N-1)=10%. It is recommended to have at least 8 vmstorage nodes in the cluster. The recommended number of vmstorage nodes should be multiplied by -replicationFactor if replication is enabled - see replication and data safety docs for details.

• Time series sharding. Received time series are consistently sharded by vminsert between configured vmstorage nodes. As a sharding key, vminsert is using time series name and labels, respecting their order. If the order of labels in a time series is constantly changing, this could cause wrong sharding calculation and result in uneven and suboptimal time series distribution across available vmstorages. It is expected that the client who is pushing metrics is responsible for consistent label order (like Prometheus or vmagent during scraping). If this can’t be guaranteed, set -sortLabels=true command-line flag to vminsert. Please note that sorting may increase CPU usage for vminsert.

• Network instability between cluster components (vminsert, vmselect, vmstorage) may lead to increased error rates, timeouts, or degraded performance. Check resource usage graphs for all components on the official Grafana dashboard for VictoriaMetrics cluster . If the graphs show high CPU usage, then the cluster is likely overloaded and requires more resources. Note that short-lived 100% CPU spikes may not be visible in metrics with typical 10–30s scrape intervals, but can still cause transient network failures. In such cases, check CPU usage at the OS level with higher-resolution tools. Consider increasing -vmstorageDialTimeout and -rpc.handshakeTimeout Available from v1.124.0 to mitigate the effects of CPU spikes.

  If resource usage appears normal but networking issues persist, the root cause is likely outside VictoriaMetrics. This may be caused by unreliable or congested network links, especially across availability zones or regions. In multi-AZ setups, consider a multi-level cluster with region-local load balancers to reduce cross-zone connections. If the network cannot be improved, increasing timeouts such as -vmstorageDialTimeout, -rpc.handshakeTimeout Available from v1.124.0 , or -search.maxQueueDuration may help, but should be done cautiously, as higher timeouts can impact cluster stability in other ways. Keep in mind that VictoriaMetrics assumes reliable networking between components. If the network is unstable, the overall cluster stability may degrade regardless of resource availability.

The obvious solution to VictoriaMetrics cluster instability is to make sure cluster components have sufficient free resources to handle the increased workload gracefully. See capacity planning docs and cluster resizing and scalability docs for details.

Too much disk space used #

If too much disk space is used by a single-node VictoriaMetrics or by vmstorage component at VictoriaMetrics cluster , then please check the following:

• Make sure that there are no old snapshots, since they can occupy disk space. See how to work with snapshots , snapshot troubleshooting and vmbackup troubleshooting .

• Under normal conditions, the size of <-storageDataPath>/indexdb folder must be smaller than the size of <-storageDataPath>/data folder, where -storageDataPath is the corresponding command-line flag value. This can be checked by the following query if VictoriaMetrics monitoring is properly set up:

        sum(vm_data_size_bytes{type=~"indexdb/.+"}) without(type)
    /
  sum(vm_data_size_bytes{type=~"(storage|indexdb)/.+"}) without(type)
      

  If this query returns values greater than 0.5, then it is likely there is a high churn rate issue, that results in excess disk space usage for both the indexdb and data folders under the -storageDataPath folder. The solution is to identify and fix the source of the high churn rate with the cardinality explorer .

Monitoring #

Having proper monitoring can help identify and prevent most of the issues listed above.

Grafana dashboards contain panels reflecting the health state, resource usage, and other specific metrics for VictoriaMetrics components.

Check the list of recommended alerting rules for VictoriaMetrics components to receive notifications about issues and receive recommendations for resolving them.

Internally, we rely heavily on both dashboards and alerts, and we constantly improve them. It is important to stay up to date with such changes.

Filesystem read corruption on ZFS #

On some ZFS filesystems, mixing reads from memory-mapped files (mmap) with usage of the mincore() syscall can trigger a bug in the ZFS in-memory cache (ARC), potentially resulting in data read corruption in VictoriaMetrics processes. This scenario has been observed when VictoriaMetrics instances access data directories on ZFS.

Symptoms: Note that the source code for the VictoriaMetrics cluster is located in the cluster branch.

• Unexpected read errors when accessing data on ZFS.
• Corrupted or inconsistent query results.
• Crashes or panics in storage/query components when reading from ZFS.

It could be mitigated with the --fs.disableMincore flag:

      ./bin/victoria-metrics --storageDataPath /path/to/zfs/data --fs.disableMincore