Manage TigerGraph services

Although the end user accesses TigerGraph by running GSQL or accessing the GraphStudio GUI, TigerGraph is actually built on many more services than just those two alone. Each has its own function and can be started or stopped individually in the case of a problem with one part, though several services rely on one another.

Consult the Glossary for definitions of the TigerGraph services.

Start, stop or restart a service

Use the command gadmin start all to start all the TigerGraph services. This is the first step when running a new TigerGraph instance. If you are using Docker, for example, you must run gadmin start all after starting the Docker container before you can use GSQL.

Use gadmin stop, gadmin start or gadmin restart with individual services such as GSQL, Kafka, or RESTPP to control them separate from the others.

The restart and stop commands prompt you for confirmation each time they are run. To avoid this, run the commands with the -y flag to bypass the confirmation.

Bypass confirmation with -y

gadmin restart gpe -y
gadmin stop all -y
gadmin start all

See the GADMIN Commmand Glossary for more details about gadmin start, stop, and restart.

Graceful Shutdown of RESTPP

When stopping RESTPP by gadmin stop restpp, the RESTPP will wait until all ongoing requests are complete before the timeout. Any long-running queries will be discarded once gadmin stop command hits the timeout. The timeout is defined by the Gadmin.StopServiceDefaultTimeoutMS configuration parameter (200s by default). The --timeout flag of gadmin stop will override the value of the configuation parameter.

Graceful Shutdown of GPE

When you stop GPE using gadmin stop gpe, the process waits for a specific duration to allow ongoing requests to complete gracefully. Any long-running queries will be discarded once the gadmin stop command hits the timeout. The timeout is defined by the GPE.StopTimeoutMS configuration parameter, which is 300 seconds (5 minutes) by default.

The graceful shutdown period is set to half of the GPE.StopTimeoutMS value. This means after 2 minutes and 30 seconds, GPE is forcefully stopped, causing any incomplete queries to fail.

Additionally, the gadmin stop command has a --timeout flag that can override the value of the GPE.StopTimeoutMS configuration parameter.

Moreover, the 5s grace period can be adjusted by adding ShutdownGracePeriodMs=<YOUR_CHOICE_OF_GRACE_PERIOD_MS>; to the GPE.BasicConfig.Env parameter.

Check service status

Use gadmin status all or gadmin status with an individual service to check its state. See Service status definitions for detailed information on the returned results.