Atomicorp

Hi folks,

I have a client who backs up to Jungle Disk and although they have other servers running ASL and Jungle Disk without a problem we have a new server which despite having a very vanilla configuration doesn't seem to allow the jungle disk server process to run properly. The /var/log/messages log file contains the following when I start the service...

Dec 21 17:09:34 london03 kernel: junglediskserve[15973]: segfault at 48 ip a5b9496e sp bfcba0d0 error 4 in libpthread-2.5.so[a5b8e000+15000]
Dec 21 17:09:34 london03 kernel: grsec: From 88.xx.xx.xx: Segmentation fault occurred at 00000048 in /usr/local/bin/junglediskserver[junglediskserve:15973] uid/euid:0/0 gid/egid:0/0, parent /sbin/init[init:1] uid/euid:0/0 gid/egid:0/0
Dec 21 17:09:34 london03 kernel: grsec: From 88.xx.xx.xx: denied resource overstep by requesting 4096 for RLIMIT_CORE against limit 0 for /usr/local/bin/junglediskserver[junglediskserve:15973] uid/euid:0/0 gid/egid:0/0, parent /sbin/init[init:1] uid/euid:0/0 gid/egid:0/0

Now I can see that this is because there is a segfault but I want to make sure that this is a problem with JD before submitting a ticket to them. Any ideas or thoughts?

The ip address is the one I am SSH'd to the box from.

Cheers

George

I recently had a problem with CouchDB not starting on the ASL kernel. I got nothing more than a "denied resource overstep by requesting 4096 for RLIMIT_CORE against limit 0" message in /var/log/messages. It turned out that I needed to run execstack -c on the Erlang beam.smp binary. Because RWXMAP_LOGGING is disabled by default I wasn't seeing the "denied RWX mmap" message, which would have helped finding the cause. See here for the thread: https://www.atomicorp.com/forum/viewtop ... f=3&t=5601

So, you could try enabling RWXMAP_LOGGING in /etc/asl/config, run 'asl -s -f' and check /var/log/messages again after trying to start the program.

Keep in mind if your kernel is locked, to turn on rwx logging will also require a reboot.

Ah ok, so after changing the default of RWXMAP_LOGGING=no to yes but without rebooting there's no change.

I'll need to schedule a reboot of the machine which should be easier over the holidays.

Thanks for the suggestion breun, and thanks for clarifying mike

I believe you'll also need to run 'asl -s -f'. (Mike, correct me if I'm wrong.)

So:

1. Change the setting in /etc/asl/config
2. Run 'asl -s -f'
3. Reboot

(Steps 1 and 2 could probably be replaced by changing the setting via the ASL web interface if you have that installed and prefer using a GUI.)

Yep, always run "asl -s -f" after you make any config changes. That sets the changes in the security policy.

And you have to reboot because the kernel was already locked at S99.

Finally got round to rebooting this machine today - all seemed to do the trick! Thanks chaps.

Sorry to dig up an old thread, but I'm having the same issue right now trying to install JungleDisk with ASL installed - thought it best to carry on here as it was #1 in Google already.

My error is the very same:
I have already tried running `execstack -c /usr/local/bin/junglediskserver` and RWXMAP_LOGGING was already set to yes.

This segfault only occurs on install (or reinstall).
Restarting the junglediskserver service just gives the following error: I have been in touch with JungleDisk and everything is in order with the account and licence.xml file.


Any ideas?

@chrismcb did you restart the box too?

ghazlewood wrote:@chrismcb did you restart the box too?

Yes, server has been rebooted.

RWXMAP_LOGGING was already enabled from a fresh install of ASL.

The funny thing is that the segfault is only on install, not every time the service is started/restarted.


Using the debug command: Doesn't bring up anything of use either

strace your application, thats always a good way to find out why an application isnt starting, no matter what kernel you are using.

strace -fF whatever_your_apps_start_command_is

Thanks... I have ran the trace and saw a few "no such file or directory" errors.

I'm wondering if, due to the segfaults happening at install, that the program hasn't been installed correctly and ASL is blocking the install - rather than the execution?

Never used the strace facility, so attached the file - in case it is clearer to anyone else.


Thanks

Just a quick follow up on this from me.

The steps to get it installed correctly were:

• Download the RPM from the Jungledisk site
• yum install junglediskserver*
• killall junglediskserver
• paxctl -c /usr/local/bin/junglediskserver
• paxctl -m /usr/local/bin/junglediskserver
• Open ports 80 and 443 outbound
• Create /etc/jungledisk/junglediskserver-licence.xml file
• service junglediskserver start

You can then proceed to configure with the remote management console.