mbettex
commented
3 years ago I investigated the issue and found out that this issue is not happening on the few environments that I have where we don't use docker, but where api-umbrella is directly installed on the vm.
In case somebody also has the issue and is interrested in a solution, for now, I apply a workaround which is to
- monitor the number of defunct children of the api-umbrella container every hour to keep track of things
- restart the api-umbrella in the docker container every day to keep things under control
I am using the following commands to monitor the number of defunct child process
apiUmbrellaPid=$(ps -aux | grep perl | grep api-umbrella | awk '{print $2}')
defunctChildCount=$(pgrep -P $apiUmbrellaPid | xargs ps -p | awk '{print $3}' | grep 'Z' | wc -l)
echo $defunctChildCountI am using the following command to restart the api-umbrella
docker exec my-apiumbrella-docker-container api-umbrella restart
Hello,
My company is using api-umbrella on docker (nrel/api-umbrella:0.15.1) on top of ubuntu (16.04.5) in multiple environments.
I noticed a strange behavior in those environments. Approximately every 2 minutes, something in the docker container of api-umbrella is creating a zombie processes.
Below is the api-umbrella docker container
Below is an extract of the processes that are currently running on one of these virtual machines.
I checked in 3 enviroments that I have this setup and the perl process in the api-umbrella container has 987, 1328 and 15465 child processes and counting.
Eventually the linux vm runs out of processes and has to be rebooted.
Any help or clue about this issue would be much appreciated.