switching from single threaded gzip to as many cores as possible using xz. We...

switching from single threaded gzip to as many cores as possible using xz. We should generate smaller log artifacts than before and get them faster.

continuousintegration/orchestration/delete_instance.sh

@@ -71,7 +71,7 @@ if [ $collectlogs == 1 ] ; then
   for podcontainer in "init -c ctainit" "client -c client" "ctacli -c ctacli" "ctaeos -c mgm" "ctafrontend -c ctafrontend" "kdc -c kdc" "tpsrv01 -c taped" "tpsrv01 -c rmcd" "tpsrv02 -c taped" "tpsrv02 -c rmcd"; do
     kubectl --namespace ${instance} logs ${podcontainer} > ${tmpdir}/$(echo ${podcontainer} | sed -e 's/ -c /-/').log
   done
-  kubectl --namespace ${instance} exec ctacli -- tar -C /mnt/logs -zcf - . > ${tmpdir}/varlog.tgz
+  kubectl --namespace ${instance} exec ctacli -- bash -c "XZ_OPT='-0 -T0' tar -C /mnt/logs -Jcf - ." > ${tmpdir}/varlog.tar.xz
 
   if [ ! -z "${CI_PIPELINE_ID}" ]; then
     # we are in the context of a CI run => save artifacts in the directory structure of the build