Re: big update killed 1 host
Ask the amanda and systemd package maintainer for centos. Since your problem is to backup the amanda server,you can try to use the "local" auth. Jean-Louis On 02/02/17 04:04 PM, Jon LaBadie wrote: > On Sun, Jan 29, 2017 at 10:45:37PM -0500, Jon LaBadie wrote: >> I added updates to my CentOS Amanda server that >> changed the OS release from 7.2 to 7.3. Amanda >> version remained at 3.3.3. Before update, all >> clients including the server were working fine. >> >> After the update, all clients backed up fine >> except the server itself. >> >> amcheck -c gives an error: >>tcpm_recv_token: Invalid size. >> >> The amcheck debug file ends with reading a string >> from a socket and saying: >> >>amcheck-clients: tcpm_recv_token: invalid size "Executing: >> /usr/sbin/amandad \ >> -auth=bsdtcp amdump amdumpd\n" >>amcheck-clients: security_stream_seterr(0x7f6be589fac0, tcpm_recv_token: \ >> invalid size: "Executing: /usr/sbin/amandad \ >> -auth=bsdtcp amdump amdumpd\n") >> >> The amandad debug file contains these lines: >> >>amandad: security_stream_seterr(0x7fcfccf15e00, write error to: \ >> Bad file descriptor) >>amandad: security_seterror(handle=0x7fcfccf14f00, driver=0x7fcfcac0f7e0 \ >> (BSDTCP) error=write error to: Bad file descriptor) >> >> The problem is neither firewall nor selinux >> related as I've turned both off and still see >> the same error. > Just a short reiteration and addition of one piece. > > - Large update of amanda server CentOS 7.2 to 7.3 (900 packages) > - Amanda went from 3.3.3-13 to 3.3.3-17 > - No amanda config changes > - Server can backup all remote clients > - Server can not backup itself > - amcheck, amdump, amrecover all fail with same error: > tcpm_recv_token: invalid size "Executing: /usr/sbin/amandad -auth=bsdtcp > amdump amdumpd\n" > > The new data piece, I downgraded to the previous amanda version, > 3.3.3-13 and amcheck fails with the same error. At this kind of > rules out the newer amanda packages as the cause of the problem, > I re-updated them to 3.3.3-17. > > So, if the amanda packages are not the problem and my amanda > configuration has not changed, where do I look? > > Jon
Re: big update killed 1 host
On Sun, Jan 29, 2017 at 10:45:37PM -0500, Jon LaBadie wrote: > I added updates to my CentOS Amanda server that > changed the OS release from 7.2 to 7.3. Amanda > version remained at 3.3.3. Before update, all > clients including the server were working fine. > > After the update, all clients backed up fine > except the server itself. > > amcheck -c gives an error: > tcpm_recv_token: Invalid size. > > The amcheck debug file ends with reading a string > from a socket and saying: > > amcheck-clients: tcpm_recv_token: invalid size "Executing: > /usr/sbin/amandad \ >-auth=bsdtcp amdump amdumpd\n" > amcheck-clients: security_stream_seterr(0x7f6be589fac0, tcpm_recv_token: \ >invalid size: "Executing: /usr/sbin/amandad \ >-auth=bsdtcp amdump amdumpd\n") > > The amandad debug file contains these lines: > > amandad: security_stream_seterr(0x7fcfccf15e00, write error to: \ >Bad file descriptor) > amandad: security_seterror(handle=0x7fcfccf14f00, driver=0x7fcfcac0f7e0 \ >(BSDTCP) error=write error to: Bad file descriptor) > > The problem is neither firewall nor selinux > related as I've turned both off and still see > the same error. Just a short reiteration and addition of one piece. - Large update of amanda server CentOS 7.2 to 7.3 (900 packages) - Amanda went from 3.3.3-13 to 3.3.3-17 - No amanda config changes - Server can backup all remote clients - Server can not backup itself - amcheck, amdump, amrecover all fail with same error: tcpm_recv_token: invalid size "Executing: /usr/sbin/amandad -auth=bsdtcp amdump amdumpd\n" The new data piece, I downgraded to the previous amanda version, 3.3.3-13 and amcheck fails with the same error. At this kind of rules out the newer amanda packages as the cause of the problem, I re-updated them to 3.3.3-17. So, if the amanda packages are not the problem and my amanda configuration has not changed, where do I look? Jon -- Jon H. LaBadie j...@jgcomp.com 11226 South Shore Rd. (703) 787-0688 (H) Reston, VA 20190 (703) 935-6720 (C)
Re: big update killed 1 host
On Mon, Jan 30, 2017 at 02:07:46PM -0500, Jean-Louis Martineau wrote: > Jon, > > Why systemd print the ExecStart on the socket? it's a systemd bug or a > misconfiguration. > > Jean-Louis > I don't know that "systemd" is printing the string. I simply said the string matches what is in the amanda@.service file. To be more precise, the debug file shows character by character the reads from the socket. It is 'g' ':' '' "the string" '' I modified the string in amanda@.service and the mods showed up in the debug file. Any ideas what might be the misconfiguration? On Mon, Jan 30, 2017 at 08:23:55PM +0100, Stefan G. Weichinger wrote: > Am 2017-01-30 um 20:07 schrieb Jean-Louis Martineau: > > Jon, > > > > Why systemd print the ExecStart on the socket? it's a systemd bug or a > > misconfiguration. > > correct, it should be in amanda@.service > > -> (on gentoo linux): > > # cat /usr/lib64/systemd/system/amanda@.service ... > # cat /usr/lib64/systemd/system/amanda.socket ... While my corresponding files vary slightly from yours, I pulled copies from a backup. They have not changed from before the updates and amanda was running fine. Jon -- Jon H. LaBadie j...@jgcomp.com 11226 South Shore Rd. (703) 787-0688 (H) Reston, VA 20190 (703) 935-6720 (C)
Re: big update killed 1 host
Am 2017-01-30 um 20:07 schrieb Jean-Louis Martineau: > Jon, > > Why systemd print the ExecStart on the socket? it's a systemd bug or a > misconfiguration. correct, it should be in amanda@.service -> (on gentoo linux): # cat /usr/lib64/systemd/system/amanda@.service [Unit] Description=Amanda Backup System After=local-fs.target [Service] User=amanda Group=amanda ExecStart=/usr/libexec/amanda/amandad -auth=bsdtcp amdump amindexd amidxtaped StandardInput=socket StandardOutput=socket # cat /usr/lib64/systemd/system/amanda.socket [Unit] Description=Amanda Socket [Socket] ListenStream=10080 Accept=true [Install] WantedBy=sockets.target
Re: big update killed 1 host
Jon, Why systemd print the ExecStart on the socket? it's a systemd bug or a misconfiguration. Jean-Louis On 30/01/17 12:38 AM, Jon LaBadie wrote: > Follow-up, > > amrecover gives the same error whether run from the server/client > or from another client. > > On the server/client I also ran: > >$ amservice mums.jgcomp.com bsdtcp noop < /dev/null >Request failed: tcpm_recv_token: invalid size: "Executing: \ >/usr/sbin/amandad -auth=bsdtcp amdump amdumpd\n" > > According to one wiki document, > >http://wiki.zmanda.com/index.php/Selfcheck_request_failed > > this suggests a failure in bsdtcp authentication. > > BTW the string "/usr/sbin/amandad -auth=bsdtcp amdump amdumpd" > matches the "ExecStart=" value in the systemd service file > "/usr/lib/systemd/system/amanda@.service". > > > On Sun, Jan 29, 2017 at 10:45:37PM -0500, Jon LaBadie wrote: >> I added updates to my CentOS Amanda server that >> changed the OS release from 7.2 to 7.3. Amanda >> version remained at 3.3.3. Before update, all >> clients including the server were working fine. >> >> After the update, all clients backed up fine >> except the server itself. >> >> amcheck -c gives an error: >>tcpm_recv_token: Invalid size. >> >> The amcheck debug file ends with reading a string >> from a socket and saying: >> >>amcheck-clients: tcpm_recv_token: invalid size "Executing: >> /usr/sbin/amandad \ >> -auth=bsdtcp amdump amdumpd\n" >>amcheck-clients: security_stream_seterr(0x7f6be589fac0, tcpm_recv_token: \ >> invalid size: "Executing: /usr/sbin/amandad \ >> -auth=bsdtcp amdump amdumpd\n") >> >> The amandad debug file contains these lines: >> >>amandad: security_stream_seterr(0x7fcfccf15e00, write error to: \ >> Bad file descriptor) >>amandad: security_seterror(handle=0x7fcfccf14f00, driver=0x7fcfcac0f7e0 \ >> (BSDTCP) error=write error to: Bad file descriptor) >> >> The problem is neither firewall nor selinux >> related as I've turned both off and still see >> the same error. >> >> Ideas? >> >> Jon >> -- >> Jon H. LaBadie j...@jgcomp.com >> 11226 South Shore Rd. (703) 787-0688 (H) >> Reston, VA 20190 (703) 935-6720 (C) End of included message <<<
Re: big update killed 1 host
Am 2017-01-30 um 06:38 schrieb Jon LaBadie: > Follow-up, > > amrecover gives the same error whether run from the server/client > or from another client. > > On the server/client I also ran: > > $ amservice mums.jgcomp.com bsdtcp noop < /dev/null > Request failed: tcpm_recv_token: invalid size: "Executing: \ > /usr/sbin/amandad -auth=bsdtcp amdump amdumpd\n" > > According to one wiki document, > > http://wiki.zmanda.com/index.php/Selfcheck_request_failed > > this suggests a failure in bsdtcp authentication. > > BTW the string "/usr/sbin/amandad -auth=bsdtcp amdump amdumpd" > matches the "ExecStart=" value in the systemd service file > "/usr/lib/systemd/system/amanda@.service". maybe I misunderstand, but on the amanda *server* you need: ExecStart=/usr/sbin/amandad -auth=bsdtcp amdump amindexd amidxtaped as far as I understand. amandad has to "provide" index services as well. I don't know of any "amdumpd" so far (?)
Re: big update killed 1 host
Follow-up, amrecover gives the same error whether run from the server/client or from another client. On the server/client I also ran: $ amservice mums.jgcomp.com bsdtcp noop < /dev/null Request failed: tcpm_recv_token: invalid size: "Executing: \ /usr/sbin/amandad -auth=bsdtcp amdump amdumpd\n" According to one wiki document, http://wiki.zmanda.com/index.php/Selfcheck_request_failed this suggests a failure in bsdtcp authentication. BTW the string "/usr/sbin/amandad -auth=bsdtcp amdump amdumpd" matches the "ExecStart=" value in the systemd service file "/usr/lib/systemd/system/amanda@.service". On Sun, Jan 29, 2017 at 10:45:37PM -0500, Jon LaBadie wrote: > I added updates to my CentOS Amanda server that > changed the OS release from 7.2 to 7.3. Amanda > version remained at 3.3.3. Before update, all > clients including the server were working fine. > > After the update, all clients backed up fine > except the server itself. > > amcheck -c gives an error: > tcpm_recv_token: Invalid size. > > The amcheck debug file ends with reading a string > from a socket and saying: > > amcheck-clients: tcpm_recv_token: invalid size "Executing: > /usr/sbin/amandad \ >-auth=bsdtcp amdump amdumpd\n" > amcheck-clients: security_stream_seterr(0x7f6be589fac0, tcpm_recv_token: \ >invalid size: "Executing: /usr/sbin/amandad \ >-auth=bsdtcp amdump amdumpd\n") > > The amandad debug file contains these lines: > > amandad: security_stream_seterr(0x7fcfccf15e00, write error to: \ >Bad file descriptor) > amandad: security_seterror(handle=0x7fcfccf14f00, driver=0x7fcfcac0f7e0 \ >(BSDTCP) error=write error to: Bad file descriptor) > > The problem is neither firewall nor selinux > related as I've turned both off and still see > the same error. > > Ideas? > > Jon > -- > Jon H. LaBadie j...@jgcomp.com > 11226 South Shore Rd. (703) 787-0688 (H) > Reston, VA 20190 (703) 935-6720 (C) >>> End of included message <<< -- Jon H. LaBadie j...@jgcomp.com 11226 South Shore Rd. (703) 787-0688 (H) Reston, VA 20190 (703) 935-6720 (C)
