[Bug 231077] New: iscsi device mount points don't get automounter on system boot
https://bugzilla.novell.com/show_bug.cgi?id=231077 Summary: iscsi device mount points don't get automounter on system boot Product: SUSE Linux 10.1 Version: Final Platform: All OS/Version: SuSE Linux 10.0 Status: NEW Severity: Critical Priority: P5 - None Component: Other AssignedTo: bnc-team-screening@forge.provo.novell.com ReportedBy: rsarraf@netapp.com QAContact: qa@suse.de CC: jballif@novell.com, xdl-novell-bugzilla@netapp.com Currently, when a system is configured with an iSCSI device to a mount point, the device is not mounted automatically at system boot. Assuming: /dev/sdc1 /data ext3 defaults,_netdev 0 0 where /dev/sdc is an iSCSI device. I've noticed that the flow of init is: 1) Execute boot.localfs. Here it tries to mount the iSCSI device but fails because that particular device won't be present there because iSCSI service is not running. 2) Bring up the network 3) Bring up the iSCSI service. Ideally, on seeing the _netdev option in /etc/fstab, the OS should hold the mount of that particular device because it depends on the network. And since it is an iSCSI device, IMO iSCSI service should run right after the network service and before any application daemon service. I also tried with the "nfsroot" option in /etc/sysconfig/network/ifcfg-eth* files but the issue is still there. I also tried with: Required-Start: boot.udev network open-iscsi but that still didn't solve the problem. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 chrubis@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|bnc-team- |ro@novell.com |screening@forge.provo.novell| |.com | -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 ro@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|ro@novell.com |hare@novell.com ------- Comment #1 from ro@novell.com 2007-01-10 09:28 MST ------- Hannes: how is the mouting of iscsi devices supposed to work ? since there is no mount command in /etc/init.d/open-iscsi, is there some functionality in the iscsi daemon to mount these devices as they get available ? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 ------- Comment #2 from rsarraf@netapp.com 2007-01-11 01:31 MST ------- I did some more finding on this bug and found that, AFAIK, SLES10 uses hotplug events instead of netfs scripts. So I changed the _netdev option in /etc/fstab to hotplug and devices were being automounted. But this still doesn't serve the purpose. The device gets mounted to the mount point only after the hotplug event is generated. And the event is generated when open-iscsi is run and the OS senses more scsi devices (LUNs being mapped to the server over iSCSI). I come up with the above statement because there are odd orders of open-iscsi in the runlevels. A freshly installed SLES10 has a correct order set for open-iscsi, S06 which is right after network (S05).This is fine. But we did notice on all of our machines, where ocfs cluster was configured, that the startup order for open-iscsi had changed to S15. Now if the change of open-iscsi from S06 to S15 was intentional by ocfs scripts, we have a problem because other applications (postfix et cetera which have a lower startup order)will fail if they use the iSCSI devices. And I believe we will have customers who would be running other applications also on the clusters. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 hare@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |hare@novell.com AssignedTo|hare@novell.com |lmb@novell.com ------- Comment #3 from hare@novell.com 2007-01-11 04:56 MST ------- This looks like something the ocfs folks did. Lars, any information about this? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 lmb@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |lmb@novell.com AssignedTo|lmb@novell.com |hare@novell.com ------- Comment #4 from lmb@novell.com 2007-01-11 05:16 MST ------- Uh? First, mere quoting of numbers doesn't help, as insserv generates them on-the-fly, so I have no idea what the _relative_ order of the scripts is. Second, ocfs2 can't make iscsi start later (relatively speaking), because it only gets to specify its own dependencies - network, syslog, named, and preferably, sshd/iscsi (if present). Which means insserv must reorder them after those, but doesn't say anything about where iscsi gets sorted to. (Hannes, you really should know that ;-) The approach using hotplug events seems correct - the filesystems get mounted as the devices appear. The issue seems to be that SLES10 is a mix between event-driven (hotplug) and static dependencies (init script deps). NFS for example also doesn't get mounted by hotplug yet, but by a dedicated nfs script. It seems that for SLES10 SP1, a possible fix is to add mount to the iscsi script, and then the iscsi dependency can be used. Ultimately, we need to switch everything to event driven, but that would require that we knew which services had which dependencies. Sigh. SLES11 material. We've got some ideas regarding that as it happens, but it won't be implemented in SLES10. Reassigning back to Hannes, as I can't do anything here. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 ------- Comment #5 from rsarraf@netapp.com 2007-01-11 06:07 MST ------- Yes, you're correct that ocfs2 has really nothing to do with it. The init scripts of ocfs2 and o2cb also just show a "Should-Start" dependency on open-iscsi. There's nothing odd in the pre/post installation scripts of the ocfs2 RPMs which would be changing the order. Doing just an installation of the ocfs2 RPMs (supplied by SLES10) doesn't create this bug. This bug is only seen (the order being changed) when the user does a configuration of cluster services using ocfs2 following the docs available at: http://oss.oracle.com/projects/ocfs2/dist/documentation/ocfs2_users_guide.pd... If ocfs2 configuration is the culprit which is modifying the boot order _internally_, it still is an ocfs2 related issue which breaks other services of the OS. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 ------- Comment #6 from lmb@novell.com 2007-01-11 06:14 MST ------- Ritesh, that's not quite right. Please see my explanation in comment #4 for the real cause. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 hare@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |rsarraf@netapp.com ------- Comment #7 from hare@novell.com 2007-01-16 01:30 MST ------- Ritesh, the actual number of the init scripts don't quite matter, the more important point is the relative order of them. So please post the listing of /etc/init.d/rc5.d before and after installing/configuring ocfs cluster. And possibly point out where you see the problem. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 rsarraf@netapp.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|rsarraf@netapp.com | ------- Comment #8 from rsarraf@netapp.com 2007-01-25 01:48 MST ------- Hi Hannes, After digging more into it I found out that it really doesn't have to do anything with ocfs2. On a freshly (cdrom) installed SLES10, where open-iscsi is selected during OS installation, the order for open-iscsi is incorrect. I've verified in on my boxes. Here's the output: linux-163:~ # ls /etc/init.d/rc5.d/ K06cupsrenice K13microcode K21acpid S02haldaemon S09powersaved K08ocfs2 K13novell-zmd K21dbus S05network S09splash K09o2cb K13powersaved K21earlysyslog S06slpd S09sshd K10open-iscsi K13splash K21fbset S06syslog S10nscd K11cron K13sshd K21irq_balancer S07auditd S10postfix K11smbfs K14nfs K21random S07portmap S10xdm K11xendomains K14nfsboot K21resmgr S07splash_early S10xend K11xinetd K15auditd S01acpid S08nfs S11cron K12nscd K15portmap S01dbus S08nfsboot S11smbfs K12postfix K15splash_early S01earlysyslog S09alsasound S11xendomains K12xdm K16slpd S01fbset S09boot.apparmor S11xinetd K12xend K16syslog S01irq_balancer S09cups S12open-iscsi K13alsasound K17network S01random S09kbd S13o2cb K13boot.apparmor K20earlykbd S01resmgr S09microcode S14ocfs2 K13cups K20haldaemon S02earlykbd S09novell-zmd S16cupsrenice All I did was to enable open-iscsi using `chkconfig open-iscsi on` Sorry for the delay. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 hare@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |ASSIGNED -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 hare@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|hare@novell.com |werner@novell.com Status|ASSIGNED |NEW ------- Comment #9 from hare@novell.com 2007-01-26 03:43 MST ------- Hmm. Appearently chkconfig is getting slightly confused here. Werner, can you help? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 werner@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|werner@novell.com |hare@novell.com ------- Comment #10 from werner@novell.com 2007-01-29 03:06 MST ------- The order is correct: [...] S11xendomains S12open-iscsi [...] and the scripts shows: ### BEGIN INIT INFO # Provides: iscsi # Required-Start: $network # Should-Start: xendomains # Required-Stop: # Should-Stop: # Default-Start: 3 5 # Default-Stop: # Short-Description: Starts and stops the iSCSI client initiator # ### END INIT INFO Next time doing an ls please set LANG=POSIX or LC_ALL=POSIX, to get a POSIX sort as the boot and rc master scripts do this in /etc/rc.status. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 ------- Comment #11 from rsarraf@netapp.com 2007-01-29 06:28 MST ------- But then how would the machine behave if my mail spool directory was on the iSCSI device? (In Comment #8, Postfix => S10 whereas open-iscsi => S12) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 hare@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |ASSIGNED ------- Comment #12 from hare@novell.com 2007-01-29 08:45 MST ------- Well, appearently Xen has to be started prior to iscsi. Sort-of makes sense if someone wants to have their Xen images on iscsi. However, this shuffles the order around. So, either you should disable xendomains (ie Xen entirely) or add an Should-Start: iscsi to /etc/init.d/postfix. Hmm. Will ask whether this is supposed to work. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 ------- Comment #13 from rsarraf@netapp.com 2007-01-29 14:19 MST ------- It might work. But expecting the _end-user_ to make modifications in the initscripts to get his application run smoothly doesn't look good, IMO. If it really is needed, Yast or chkconfig should be doing that. But again, How ? How would Yast or chkconfig sense that postfix (or any user application) has its mail spool on the iSCSI device ? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 hare@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |rsarraf@netapp.com ------- Comment #14 from hare@novell.com 2007-01-31 01:36 MST ------- Right, found the culprit. It's actually the line Should-Start: xendomains in /etc/init.d/open-iscsi. It got added due to bug #177381, but even then it wasn't quite clear whether it was the right solution. So we should remove that line in for the open-iscsi rpm and have the Xen folks come up with a better solution. Is it sufficient to have this change for SLES10 SP1 or do you require a fix for SLES10 GA, too? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 rsarraf@netapp.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|rsarraf@netapp.com | ------- Comment #15 from rsarraf@netapp.com 2007-02-07 05:36 MST ------- Great thanks. That works. I'd recommend this fix to be released for SLES10 GA also, because it can affect/break application startup. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 hare@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |ihno@novell.com ------- Comment #16 from hare@novell.com 2007-02-09 07:02 MST ------- Can I get a SWAMP-ID, please? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231077 ihno@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|ihno@novell.com | ------- Comment #17 from ihno@novell.com 2007-02-13 03:10 MST ------- Is it neccessary to release this seperate or is it sufficiant to release this with SP1? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
participants (1)
-
bugzilla_noreply＠novell.com