Change Details

The loader pods should use local hypervisor storage to avoid an unnecessary load on ceph when the data is offloaded on the disk during the parsing of big repositories. Currently, on the current "classical" workers, the temporary file is written in a memory fs and the servers are configured to swap on a partition hosted in the local storage. The same mechanism can't be used for the kubernetes infrastructure as the swap must be disabled for the nodes themselves and the notion of swap doesn't exist in a pod.

The loader pods should use local hypervisor storage to avoid an unnecessary load on ceph when the data is offloaded on the disk during the parsing of big repositories. Currently, on the current "classical" workers, the temporary file is written in a memory fs and the servers are configured to swap on a partition hosted on the local storage. The same mechanism can't be used for the kubernetes infrastructure as the swap must be disabled for the nodes themselves and the notion of swap doesn't exist in a pod.

The loader pods should use local hypervisor storage to avoid an unnecessary load on ceph when the data is offloaded on the disk during the parsing of big repositories. Currently, on the current "classical" workers, the temporary file is written in a memory fs and the servers are configured to swap on a partition hosted ion the local storage. The same mechanism can't be used for the kubernetes infrastructure as the swap must be disabled for the nodes themselves and the notion of swap doesn't exist in a pod.