From Bugzilla Helper: User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.5) Gecko/20031007 Description of problem: With the new version of autofs included in RHEL 3.0-U3, automounted filesystems inexplicably become unavailable. It appears that the mount is in some sort of weird limbo, where some parts of the system believe that the filesystem is mounted, but other parts of the filesystem believe it isn't. When the automount is in this limbo state, attempts to access the filesystem generate messages in /var/log/messages such as the following: Sep 17 10:19:53 lnxtst04 automount[15355]: BUG: /apps/vendor/bin already mounted The problem can be resolved by issuing a manual umount command; subsequent attempts to access the automount filesystem will cause it to successfully be automounted and become available. The following shows a real-world example of this problem occuring (/apps/vendor/bin is an automounted filesystem): % ls /apps/vendor/bin ./ ../ (lnxtst04 : root : /var/log) % df -k /apps/vendor/bin Filesystem 1K-blocks Used Available Use% Mounted on ahfil03:/vol/vol0_03/software/vendor/bin 0 0 0 - /apps/vendor/bin (lnxtst04 : root : /var/log) % grep "/apps/vendor/bin" /proc/mounts (lnxtst04 : root : /var/log) % umount /apps/vendor/bin umount: /apps/vendor/bin: not mounted (lnxtst04 : root : /var/log) % ls /apps/vendor/bin ./ bstop@ hpgltomif@ jarsigner@ kwinsight@ matlab* rose* tmsbrowser@ ../ bsub@ iaftomif@ java* kwinspect@ mmltomif@ rose2000@ trofftomif@ .snapshot/ cleartool* idu@ javac@ kwjava@ mozilla* rose2001@ viewer@ REPnewconfname@ codewizard* igestomif@ javadoc@ kwlinker@ netscape@ rose2002@ win2k* acroexch* cwmfs@ iiusers@ javah@ lbrun* ntperl* rose2003@ win2k.pl* acroread* distill@ ileaf* javap@ lbrun.bsub* picttomif@ rose98i@ winnt@ appletviewer@ doors* ileaf5@ jdb@ lshosts@ pl2ps@ rulewizard@ wordtomif@ bbot@ dxftomif@ ileaf6@ keytool@ lsid@ policytool@ smallxd@ wptomif@ bhist@ edgcfe@ ileaf7@ kmc@ lsinfo@ pslic@ tau* xdesigner* bhosts@ flexelint* ileaffilt@ kwadmin@ lsload@ psrcdump@ tau4.2cc@ xwd8to1@ bjobs@ fmbatch@ ins_pp@ kwcc* maker* ptk* tau4.4@ bkill@ fmclient@ insra@ kwcreatemodel@ maker5@ qsilver@ tau4.4.4@ bqueues@ fminvert@ is1tomif@ kwcreatesembase@ maker6@ reader@ tau4.5@ bresume@ fmprint@ jar@ kwinforce@ maker7@ rm8tomif@ teamplayer* (lnxtst04 : root : /var/log) % df -k /apps/vendor/bin Filesystem 1K-blocks Used Available Use% Mounted on ahfil03:/vol/vol0_03/software/vendor/bin 36700160 18846240 17853920 52% /apps/vendor/bin (lnxtst04 : root : /var/log) % grep "/apps/vendor/bin" /proc/mounts ahfil03:/vol/vol0_03/software/vendor/bin /apps/vendor/bin nfs ro,v3,rsize=32768,wsize=32768,hard,intr,udp,lock,addr=ahfil03 0 0 Version-Release number of selected component (if applicable): autofs-4.1.3-12 How reproducible: Sometimes Steps to Reproduce: 1. Access an automounted filesystem 2. Wait -- the problem will eventually occur Actual Results: See the description above. Expected Results: The automounted filesystem should remain available. Additional info:
What is the version of the kernel you are using? Please also post the output of lsmod. Thanks!
All components on the test system where we're seeing this problem are vanilla RHEL 3.0-U3. Kernel version = 2.4.21-20.ELsmp % lsmod Module Size Used by Tainted: PF sr_mod 17784 0 (autoclean) (unused) ide-cd 34016 0 (autoclean) cdrom 32544 0 (autoclean) [sr_mod ide-cd] vmnet 25376 4 vmmon 42996 14 nfs 99376 10 (autoclean) lockd 58992 1 (autoclean) [nfs] mvfs 309056 119 vnode 76116 119 [mvfs] sunrpc 89212 1 [nfs lockd vnode] autofs 13620 7 (autoclean) audit 90552 3 tg3 65832 1 floppy 57488 0 (autoclean) sg 37228 0 (autoclean) microcode 6848 0 (autoclean) keybdev 2976 0 (unused) mousedev 5624 1 hid 22276 0 (unused) input 6144 0 [keybdev mousedev hid] usb-ohci 23176 0 (unused) usbcore 81056 1 [hid usb-ohci] ext3 89960 2 jbd 55060 2 [ext3] megaraid 30604 3 sd_mod 13712 6 scsi_mod 114344 4 [sr_mod sg megaraid sd_mod]
You aren't using the autofs4 kernel module. Please do the following: service autofs stop rmmod autofs service autofs start After that, an lsmod *should* show autofs4. If it doesn't, you can add the following line to your /etc/modules.conf: alias autofs autofs4 Please try this and let me know if the problem persists. Thank you.
Ah-hah. The /etc/rc.d/init.d/autofs script for autofs v4 had significant updates to it. We use a customized version of the autofs init script to compute and set localoptions: Using the old autofs (v3) init script loads the wrong autofs module. Looks like we'll have to edit our customizations into the new autofs init script.
Does running the autofs4 kernel module fix your problem? You don't specifically say it does. ;) I'm going to keep this open until I get confirmation. If this is a bug, we need to address it. Also, what are the custom changes that you maintain? Perhaps we can get them put into the rpm or upstream if they are applicable to others. -Jeff
:) To clarify: With the normal autofs-4.1.3-12 init script, the autofs4 kernel module is properly loaded instead of the autofs kernel module. With the autofs4 kernel module, we no longer experience the problem described initially within this bug. Regarding our customizations: A couple of our customizations are to force our Linux systems to ignore certain legacy NIS automount maps -- these are pretty specific to us (and we may not even need them any more; I'll need to check). The more important customization is that we set the "localoptions" variable ("LOCALOPTIONS" in the v4 script) so that we can mount different filesystems depending on the OS and OS version. Looking at the autofs v4 init script, though, it looks like an improvement has already been made here: LOCALOPTIONS can now be set in /etc/sysconfig/autofs, which means we can maintain that without modifying the autofs init script itself.
Okay, then I'll close this bug. Thank you for clarifying. As a side note, there is a feature request for something that sounds like it may work for you. Please take a look at bz #91559: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=91559 If that sounds like what you need, please let me know. Multiple people requesting a feature bumps it up on my list. Thanks!