systemd service: restart less agressively but still quicker

If a service fails to start because it's not properly configured (e.g.
directly after installation) the existing settings make the daemons
restart in a loop. Each start logs several lines to the journal
which quickly results in older but probably still relevant logs of other
serives being rotated out.

Dropping RestartSec makes systemd use the default of 100ms. So a restart
happens quicker. With dropping StartLimitInterval=0 systemd applies some
ratelimiting and (by default) only starts the service 5 times per 10s.

Replacing Restart=on-failure by Restart=on-abnormal makes systemd not
restart the service if the daemon process exits with a non-zero exit
code. It's still restarted when the process exits by a signal (other
than SIGHUP, SIGINT, SIGTERM or SIGPIPE).

Debian-Bug: https://bugs.debian.org/1055529

pdns/dnsdistdist/dnsdist.service.in

@@ -13,10 +13,8 @@ User=@service_user@
 Group=@service_group@
 SyslogIdentifier=dnsdist
 Type=notify
-Restart=on-failure
-RestartSec=2
+Restart=on-abnormal
 TimeoutStopSec=5
-StartLimitInterval=0
 
 # Tuning
 TasksMax=8192

pdns/ixfrdist.service.in

@@ -9,9 +9,7 @@ After=network-online.target time-sync.target
 [Service]
 Type=simple
 ExecStart=@bindir@/ixfrdist
-Restart=on-failure
-RestartSec=1
-StartLimitInterval=0
+Restart=on-abnormal
 
 # Sandboxing
 CapabilityBoundingSet=CAP_NET_BIND_SERVICE CAP_SETGID CAP_SETUID

pdns/pdns.service.in

@@ -11,9 +11,7 @@ SyslogIdentifier=pdns_server
 User=@service_user@
 Group=@service_group@
 Type=notify
-Restart=on-failure
-RestartSec=1
-StartLimitInterval=0
+Restart=on-abnormal
 RuntimeDirectory=pdns
 
 # Sandboxing

pdns/recursordist/pdns-recursor.service.in

@@ -10,8 +10,7 @@ ExecStart=@sbindir@/pdns_recursor --daemon=no --write-pid=no --disable-syslog --
 User=@service_user@
 Group=@service_group@
 Type=notify
-Restart=on-failure
-StartLimitInterval=0
+Restart=on-abnormal
 RuntimeDirectory=pdns-recursor
 SyslogIdentifier=pdns-recursor