ZFS and selfcheck failures

[btm1001]

I'm new to Amanda, and have been able to get basic backups going, but am stuck trying to make ZFS work.

I'm on FreeBSD 8.0 stable, and using the latest Amanda built from source. I have the system hooked up to a 24 slot changer with an LTO2 tape drive.

I edited the amanda.conf file and added :

define application-tool amzfs_sendrecv {
comment "amzfs-sendrecv"
plugin "amzfs-sendrecv"
property "DF-PATH" "/bin/df"
property "KEEP-SNAPSHOT" "YES"
property "ZFS-PATH" "/sbin/zfs"
#property "PFEXEC-PATH" "/usr/local/bin/sudo"
#property "PFEXEC" "YES"
}

define dumptype user-zfs-sendrecv {
program "APPLICATION"
application "amzfs_sendrecv"
}

I added ZFS permissions for the amanda user (operator):

# zfs allow scratch
-------------------------------------------------------------
Local+Descendent permissions on (scratch)
user operator create,destroy,mount,receive,rename,send,snapshot
-------------------------------------------------------------

and my disklist looks like:

#ns1.eng.sldomain.com /tmp comp-user-tar
ns1.eng.sldomain.com scratch user-zfs-sendrecv

When I run "amcheck DailySet1", I get the error:

Amanda Backup Client Hosts Check
--------------------------------
WARNING: ns1.<FQDN>: selfcheck request failed: timeout waiting for ACK
Client check: 1 host checked in 30.152 seconds. 1 problem found.

If I comment out the ZFS portion and just to the comp-user-tar line, amcheck finished fine.

I have verified that the operator user can create a snapshot.

I know the general error is due to a permissions/ownership issue, but I can't seem to spot WHICH file/command/directory is not set properly. Can anyone point me to some spots to check?

[martineau]

What's in the client debug files?

[btm1001]

After clearing out the logdir, I re-rean amcheck DailySet1, it failed with the same error. There were 2 logfiles reated, a device one which seems OK, and a /tmp/amanda/server/DailySet1/amcheck.20101108135240. No client files were made. This should be set up for basic BSDTCP authentication

cat amcheck.20101108135240.debug
Mon Nov 8 13:52:40 2010: amcheck: pid 86102 ruid 0 euid 2 version 3.2.0: start at Mon Nov 8 13:52:40 2010
Mon Nov 8 13:52:40 2010: amcheck: pid 86102 ruid 0 euid 2 version 3.2.0: rename at Mon Nov 8 13:52:40 2010
Mon Nov 8 13:52:40 2010: amcheck-clients: security_getdriver(name=BSD) returns 0x800b154e0
Mon Nov 8 13:52:40 2010: amcheck-clients: security_handleinit(handle=0x8022373c0, driver=0x800b154e0 (BSD))
Mon Nov 8 13:52:40 2010: amcheck-clients: dgram_bind: setting up a socket with family 2
Mon Nov 8 13:52:40 2010: amcheck-server: Spawning "/usr/local/libexec/amanda/amcheck-device /usr/local/libexec/amanda/amcheck-device DailySet1" in pipeline
Mon Nov 8 13:52:40 2010: amcheck-clients: bind_portrange2: Try port 880: Available - Success
Mon Nov 8 13:52:40 2010: amcheck-clients: dgram_bind: socket 4 bound to 0.0.0.0.880
Mon Nov 8 13:52:40 2010: amcheck-clients: dgram_send_addr(addr=0x802237400, dgram=0x800b23768)
Mon Nov 8 13:52:40 2010: amcheck-clients: (sockaddr_in *)0x802237400 = { 2, 10080, 10.1.0.9 }
Mon Nov 8 13:52:40 2010: amcheck-clients: dgram_send_addr: 0x800b23768->socket = 4
Mon Nov 8 13:52:46 2010: amcheck-server: userbad: 0
Mon Nov 8 13:52:46 2010: amcheck-server: confbad: 0
Mon Nov 8 13:52:46 2010: amcheck-server: tapebad: 0
Mon Nov 8 13:52:46 2010: amcheck-server: disklow: 0
Mon Nov 8 13:52:46 2010: amcheck-server: logbad: 0
Mon Nov 8 13:52:46 2010: amcheck-server: infobad: 0
Mon Nov 8 13:52:46 2010: amcheck-server: indexbad: 0
Mon Nov 8 13:52:46 2010: amcheck-server: pgmbad: 0
Mon Nov 8 13:52:50 2010: amcheck-clients: dgram_send_addr(addr=0x802237400, dgram=0x800b23768)
Mon Nov 8 13:52:50 2010: amcheck-clients: (sockaddr_in *)0x802237400 = { 2, 10080, 10.1.0.9 }
Mon Nov 8 13:52:50 2010: amcheck-clients: dgram_send_addr: 0x800b23768->socket = 4
Mon Nov 8 13:53:00 2010: amcheck-clients: dgram_send_addr(addr=0x802237400, dgram=0x800b23768)
Mon Nov 8 13:53:00 2010: amcheck-clients: (sockaddr_in *)0x802237400 = { 2, 10080, 10.1.0.9 }
Mon Nov 8 13:53:00 2010: amcheck-clients: dgram_send_addr: 0x800b23768->socket = 4
Mon Nov 8 13:53:11 2010: amcheck-clients: security_seterror(handle=0x8022373c0, driver=0x800b154e0 (BSD) error=timeout waiting for ACK)
Mon Nov 8 13:53:11 2010: amcheck-clients: security_close(handle=0x8022373c0, driver=0x800b154e0 (BSD))
Mon Nov 8 13:53:11 2010: amcheck: pid 86102 finish time Mon Nov 8 13:53:11 2010

[martineau]

On the third lines:
Mon Nov 8 13:52:40 2010: amcheck-clients: security_getdriver(name=BSD) returns 0x800b154e0

It state 'BSD' auth, fix the disklist/dumptype.

[btm1001]

Thanks!
changed dumptype to have the auth "bsdtcp" setting and amcheck looks to be completing properly now. I'm off to run a test backup, but this looks like it solves the problem.

define dumptype user-zfs-sendrecv {
program "APPLICATION"
application "amzfs_sendrecv"
auth "bsdtcp"
}

Can you point me to where you get the info about the return codes (such as the 0x800b154e0) you pointed out? I'd like to continue my studying.

[martineau]

0x800b154e0 is a memory address, nothing useful.

(name=BSD) is told that it is using the bsd auth, not bsdtcp like you said.