elo, It is not always reproducible, but in virtualized environment under heavy load you can see at system startup: *start-stop-daemon: did not create a valid pid in `/var/run/iscsid.pid`. Usually this helps (file located in /etc/init.d): finlandia-xen working # diff -Naur iscsid iscsid.old --- iscsid 2014-02-21 13:30:43.000000000 +0100 +++ iscsid.old 2014-02-21 14:50:53.303125000 +0100 @@ -5,7 +5,7 @@ command="/usr/sbin/iscsid" command_args="${OPTS}" -start_stop_daemon_args="-w 10" # iscsid might fail e.g. when the iSCSI kernel modules aren't available +start_stop_daemon_args="-w 100" # iscsid might fail e.g. when the iSCSI kernel modules aren't available pidfile=${PIDFILE:-/var/run/${SVCNAME}.pid} extra_started_commands="starttargets stoptargets" Reporting this just to help someone in search for this particular problem.
Why does it wait at all? Timeouts are inherently broken in this case.
(In reply to Jeroen Roovers from comment #1) > Why does it wait at all? Timeouts are inherently broken in this case. Yup, we are trying to boot faster and faster... so yes they are :) I am using open-iscsi-2.0.873 because previous ebuild open-iscsi-2.0.872-r2 insisted that some kernel parts should be modules (and I want monolithic one), namely: Open-iscsi will not start when iSCSI Initiator over TCP/IP is built-in, so you have to make it a module for now. Open-iscsi will not start when iSCSI Transport Attributes is built-in, so you have to make it a module for now. (see: http://etherboot.org/wiki/sanboot/gentoo_iscsi) I'm using Linux/x86_64 3.7.0-hardened and these are CONFIG_ISCSI_TCP & CONFIG_SCSI_ISCSI_ATTRS. open-iscsi-2.0.873 doesn't try to load modules so I can cleanly startup and shutdown my machine.
I ran into the same issue once I upgraded to 2.0.873. With the default wait time I get: Apr 27 11:00:46 94123 iscsid: iSCSI logger with pid=1197 started! Apr 27 11:00:46 94123 /etc/init.d/iscsid[1193]: start-stop-daemon: did not create a valid pid in `/var/run/iscsid.pid' Apr 27 11:00:46 94123 /etc/init.d/iscsid[1181]: ERROR: iscsid failed to start Apr 27 11:00:46 94123 kernel: Loading iSCSI transport class v2.0-870. Apr 27 11:00:46 94123 wpa_cli: interface wlp4s0 CONNECTED Apr 27 11:00:47 94123 kernel: ip_tables: (C) 2000-2006 Netfilter Core Team Apr 27 11:00:47 94123 kernel: nf_conntrack version 0.5.0 (16384 buckets, 65536 max) Apr 27 11:00:47 94123 iscsid: iSCSI daemon with pid=1198 started! So a race occurs which start-stop-daemon wins. Maybe the computer is just too slow :-) When increasing the wait time to 0.1 seconds as suggested by Krzysztof, the problem disappears: Apr 27 11:24:38 94123 iscsid: iSCSI logger with pid=1222 started! Apr 27 11:24:38 94123 kernel: Loading iSCSI transport class v2.0-870. Apr 27 11:24:38 94123 kernel: ip_tables: (C) 2000-2006 Netfilter Core Team Apr 27 11:24:38 94123 kernel: nf_conntrack version 0.5.0 (16384 buckets, 65536 max) Apr 27 11:24:39 94123 iscsid: iSCSI daemon with pid=1223 started!
fixed in 2.0.873