Administration troubleshooting

Docker Toolbox on Windows: New state of 'nil' is invalid

If you are using Docker Toolbox on Windows to run Sourcegraph, you may see an error in the frontend log output:

frontend |
     frontend |
     frontend |
     frontend |     New state of 'nil' is invalid.

After this error, no more frontend log output is printed.

This problem is caused by docker/toolbox#695 in Docker Toolbox on Windows. To work around it, set the environment variable LOGO=false, as in:

docker container run -e LOGO=false ... sourcegraph/server

See sourcegraph/sourcegraph#398 for more information.

Submitting a metrics dump

If you encounter performance or instability issues with Sourcegraph, we may ask you to submit a metrics dump to us. This allows us to inspect the performance and health of various parts of your Sourcegraph instance in the past and can often be the most effective way for us to identify the cause of your issue.

The metrics dump includes non-sensitive aggregate statistics of Sourcegraph like CPU & memory usage, number of successful and error requests between Sourcegraph services, and more. It does NOT contain sensitive information like code, repository names, user names, etc.

Docker Compose deployments

To create a metrics dump from a docker-compose deployment, follow these steps:

  • Open a shell to the running prometheus container:
docker exec -it prometheus /bin/sh
  • Inside the container bash shell trigger the creation of a Prometheus snapshot:
wget --post-data "" http://localhost:9090/api/v1/admin/tsdb/snapshot
  • Find the created snapshot’s name:
ls /prometheus/snapshots/
  • Tar up the created snapshot
cd /prometheus/snapshots && tar -czvf /tmp/sourcegraph-metrics-dump.tgz <snapshot-name>
  • Switch back to local shell (exit) and copy the metrics dump file to the host machine:
docker cp prometheus:/tmp/sourcegraph-metrics-dump.tgz sourcegraph-metrics-dump.tgz

Please then upload the sourcegraph-metrics-dump.tgz file to Sourcegraph support so we can inspect it.

Single-container sourcegraph/server deployments

To create a metrics dump from a single-container sourcegraph/server deployment, follow these steps:

  • Open a shell to the running container:
    1. Run docker ps to get the name of the Sourcegraph server container.
    2. Run docker exec -it <container name> /bin/bash to open a bash shell.
  • Inside the container bash shell trigger the creation of a Prometheus snapshot:
wget --post-data "" http://localhost:9090/api/v1/admin/tsdb/snapshot
  • Tar up the created snapshot
cd ~/.sourcegraph/data/prometheus/snapshots && tar -czvf /tmp/sourcegraph-metrics-dump.tgz <snapshot-name>
  • If needed, you can download the metrics dump to your local machine (current directory) using scp:
scp -r [email protected]:/tmp/sourcegraph-metrics-dump.tgz .

Please then upload the sourcegraph-metrics-dump.tgz for Sourcegraph support to access it. If desired, we can send you a shared private Google Drive folder for the upload as it can sometimes be a few gigabytes.

Kubernetes deployments

If you’re using the Kubernetes cluster deployment option,
you can create a metrics dump as follows:

  • Open a shell to the running container

    1. Run kubectl get pods to get the name of the Prometheus pod.
    2. Run kubectl exec -it <pod-name> -- /bin/bash.
  • Inside the container bash shell trigger the creation of a Prometheus snapshot:

wget --post-data "" http://localhost:9090/api/v1/admin/tsdb/snapshot
  • Tar up the created snapshot
cd /prometheus/snapshots && tar -czvf /tmp/sourcegraph-metrics-dump.tgz <snapshot-name>
  • Switch back to local shell and copy the metrics dump file over:
kubectl cp <podname>:/tmp/sourcegraph-metrics-dump.tgz /tmp/sourcegraph-metrics-dump.tgz

Again please then upload the sourcegraph-metrics-dump.tgz for Sourcegraph support to access it.

Generating pprof profiles

Please follow these instructions to generate pprof profiles.

Sourcegraph is not returning results from a repository unless “repo:” is included

If you can get repository results when you explicitly include repo:{your repository} in your search, but don’t see any results from that repository when you don’t, there are a few possible causes:

  • The repository is a fork repository (excluded from search results by default) and fork:yes is not specified in the search query.
  • The repository is an archived repository (excluded from search results by default) and archived:yes is not specified in the search query.
  • Your site config file does not include "search.index.enabled": true. It should be included, and you should set it to true; if it’s false, it means Sourcegraph won’t index anything and will only search in real-time.
  • There is an issue indexing the repository: check the logs of repo-updater and/or search-indexer.
  • The search index is unavailable for some reason: try the search query repo:<the_repository> index:only. If it returns no results, the repository has not been indexed.