Re: Failure on W2k client
On Apr 3, 2001, David Lloyd [EMAIL PROTECTED] wrote: Again: THIS IS NOT AMANDA'S FAULT! IT IS SMBCLIENT'S FAULT! You obviously cannot understand what I am saying, or you have not followed this thread. If you cannot understand English I will have this translated into any other language that I have a character set for: SMBCLIENT CAN CONTACT MY WINDOWS 2000 SERVER. IT DOES NOT FAIL. IT WORKS VIA BROADCAST AND AMANDA FAILS TO SEE THAT IT WORKS. Or to reword this: SAMBA IS ABLE TO CONTACT MY WINDOWS 2000 SERVER; SMBCLIENT CAN CONTACT IT. HOWEVER, IT DOES THIS VIA BROADCAST AND AMANDA FAILS. Amanda just runs smbclient. If Amanda says the backup failed, it's because smbclient said so. Perhaps you're running one version of smbclient by hand, but Amanda is running another, misconfigured version? Perhaps you're passing some command-line flag to smbclient that Amanda doesn't pass? Whatever the case is, it's smbclient that is failing to contact the Windows machine. Amanda plays absolutely no role here other than telling smbclient to create a tar-file off the Windows box. So how is it smbclient's fault when smbclient works? I can't understand your English... Perhaps I don't speak or write English well enough to get the message through. It can't possibly be Amanda's fault. Get the exact command line that Amanda runs (they're logged in /tmp/amanda/*.debug) and try them. If they fail to work, you'll have more material to investigate the problem. If they work, then you probably have a timing problem (such as having the backup account forbidden from logging in at the time of the backup or something). -- Alexandre Oliva Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/ Red Hat GCC Developer aoliva@{cygnus.com, redhat.com} CS PhD student at IC-Unicampoliva@{lsd.ic.unicamp.br, gnu.org} Free Software Evangelist*Please* write to mailing lists, not to me
Re: Failure on W2k client
Ok, I think finally understand what Marty is trying to say. Marty, there are many people on this list who have been using samba for years and understand it very well and how it interacts with amanda, and if we couldn't understand the gist of your initial question, don't blame our lack of understanding of the english language. :-) Marty says his samba setup works, but because of the fallback from wins to broadcast the lookup takes longer than amanda is willing to wait. Amanda times out and as far as amanda is concerned, smbclient failed. I'm not even sure what the implications of this are or how to make it work more efficiently, I'm just interpreting what he stated. Marty, is this correct? Now, this said, I still think the problem is with samba. You should 'fix' your samba setup so that it doesn't take so long to do a lookup. Todd On Wed, 4 Apr 2001, David Lloyd wrote: Marty! I suggest you put your asbestos, flame resistant suit on... Again: THIS IS NOT AMANDA'S FAULT! IT IS SMBCLIENT'S FAULT! You obviously cannot understand what I am saying, or you have not followed this thread. If you cannot understand English I will have this translated into any other language that I have a character set for: SMBCLIENT CAN CONTACT MY WINDOWS 2000 SERVER. IT DOES NOT FAIL. IT WORKS VIA BROADCAST AND AMANDA FAILS TO SEE THAT IT WORKS. Or to reword this: SAMBA IS ABLE TO CONTACT MY WINDOWS 2000 SERVER; SMBCLIENT CAN CONTACT IT. HOWEVER, IT DOES THIS VIA BROADCAST AND AMANDA FAILS. As far as I'm concerned SMB is working correctly. It is setup to use wins, wins fails and it therefore contacts my Windows 2000 server by broadcast and connects. Perfectly. So how is it smbclient's fault when smbclient works? I can't understand your English... DSL -- Todd Pfaff \ Email: [EMAIL PROTECTED] Computing and Information Services \ Voice: (905) 525-9140 x22920 ABB 132 \ FAX: (905) 528-3773 McMaster University \ Hamilton, Ontario, Canada L8S 4M1 \
Re: Failure on W2k client
My last message should have been directed to David Lloyd, not Marty Shannon. Sorry Marty! -- Todd Pfaff \ Email: [EMAIL PROTECTED] Computing and Information Services \ Voice: (905) 525-9140 x22920 ABB 132 \ FAX: (905) 528-3773 McMaster University \ Hamilton, Ontario, Canada L8S 4M1 \
Re: Failure on W2k client
Hi folks, Both amanda:sendsize and smbclient are OK, it is the interaction that is the problem. This started with my question re: read_socket_with_timeout errors in sendsize.debug. The error is generated from lib/util_sock.c in samba code. From the logs smbclient is noting a failed open to amanda:sendsize. Smbclient takes it own error message as an indicator to use the next address resolution scheme and continues to attempt to connect. Sendsize has noted the socket open failure and declared "session setup failed: code 0"; since the smbclient process hasn't died yet, sendsize continues to log information from the smbclient it spawned. A few things could happen here. Sendsize could kill and respawn an error-reporting smbclient, or wait to call the session off until smbclient exits; Smbclient could march through the host resolution options before reporting a failed open, thus not confusing amanda:sendsize. The reason, probably, that a "wins" first host lookup works for is that the smbclient file descriptor is not a socket and a different set of routines are used to perform the open, which for what ever reason are more robust or less verbose, thereby not causing sendsize to make determination that the session failed. We have a choice of where to make the fix. Changing amanda:sendsize to wait for smbclient to exit before declaring a failed session would make amanda more robust; changing smbclient to wait until the full gamut of host resolution options have been tried before reporting errors would also fix this problem from an amanda perspective but may cause other smb-dependent apps to suffer. Having amanda handle the different possiblities smbclient presents seems a good way to go.. Does anyone careto look at the code? sendsize.c line 504, maybe... We have a tentative thesis that the initial problem may be caused by the NIC being put to sleep in "powersaving" mode during inactive times (night) which is why we can't re-produce the problem during the day. We did uncheck the bit that does this in the 3Com w2k driver but didn't reboot the system. We saw no change last night so we'll try rebooting the sucker this afternoon. thanks, hurf -- Todd Pfaff wrote: Ok, I think finally understand what Marty is trying to say. Marty, there are many people on this list who have been using samba for years and understand it very well and how it interacts with amanda, and if we couldn't understand the gist of your initial question, don't blame our lack of understanding of the english language. :-) Marty says his samba setup works, but because of the fallback from wins to broadcast the lookup takes longer than amanda is willing to wait. Amanda times out and as far as amanda is concerned, smbclient failed. I'm not even sure what the implications of this are or how to make it work more efficiently, I'm just interpreting what he stated. Marty, is this correct? Now, this said, I still think the problem is with samba. You should 'fix' your samba setup so that it doesn't take so long to do a lookup. Todd On Wed, 4 Apr 2001, David Lloyd wrote: Marty! I suggest you put your asbestos, flame resistant suit on... Again: THIS IS NOT AMANDA'S FAULT! IT IS SMBCLIENT'S FAULT! You obviously cannot understand what I am saying, or you have not followed this thread. If you cannot understand English I will have this translated into any other language that I have a character set for: SMBCLIENT CAN CONTACT MY WINDOWS 2000 SERVER. IT DOES NOT FAIL. IT WORKS VIA BROADCAST AND AMANDA FAILS TO SEE THAT IT WORKS. Or to reword this: SAMBA IS ABLE TO CONTACT MY WINDOWS 2000 SERVER; SMBCLIENT CAN CONTACT IT. HOWEVER, IT DOES THIS VIA BROADCAST AND AMANDA FAILS. As far as I'm concerned SMB is working correctly. It is setup to use wins, wins fails and it therefore contacts my Windows 2000 server by broadcast and connects. Perfectly. So how is it smbclient's fault when smbclient works? I can't understand your English... DSL -- Todd Pfaff \ Email: [EMAIL PROTECTED] Computing and Information Services \ Voice: (905) 525-9140 x22920 ABB 132 \ FAX: (905) 528-3773 McMaster University \ Hamilton, Ontario, Canada L8S 4M1 \ -- Hurf Sheldon Dir. Research Systems Program of Computer Graphics 580 Rhodes Hall, Hoy Rd. Cornell University Ithaca, N.Y. 14853 voice:607 255 6713 fax:607 255 0806 email: [EMAIL PROTECTED] http://www.graphics.cornell.edu/~hurf/
Re: Failure on W2k client
Todd! Marty says his samba setup works, but because of the fallback from wins to broadcast the lookup takes longer than amanda is willing to wait. Amanda times out and as far as amanda is concerned, smbclient failed. I'm not even sure what the implications of this are or how to make it work more efficiently, I'm just interpreting what he stated. Marty, is this correct? That is correct. However I said it, not Marty as you've already noted. All I'm saying (in my own obtuse way) is that it appears "weird" that amanda/smbclient don't always get along with each other. At the moment it's going to be easier to work around this "smbclient" behaviour because this version of smbclient is going to be in production for a reasonable time yet. Simply pointing out that "it's Samba's fault" isn't at all helpful even if it is true. Telling someone that "it's in the FAQ" when they've taken the time to attempt to help someone on the list isn't at all helpful even if the answer is in the FAQ. If you're going to tell people that it's in the FAQ or to read a manual you really should post a URL or some sort of indicator where the FAQ or manual is... I don't think I need to continue this thread any further; I'll just go along and help out where I can. If some people notice that I'm repeating the FAQ at http://www.amanda.org/ then so be it. They can stay silent... DSL -- There's a sad face in the mirror And I'm sad to say it's me Like a ghost up in the attic Only love can set met free...
Re: Failure on W2k client
On Apr 4, 2001, Hurf Sheldon [EMAIL PROTECTED] wrote: Sendsize has noted the socket open failure and declared "session setup failed: code 0"; since the smbclient process hasn't died yet, sendsize continues to log information from the smbclient it spawned. And it keeps looking for the size line, AFAICT. So, whenever smbclient manages to connect to the client, it should work. We have a tentative thesis that the initial problem may be caused by the NIC being put to sleep in "powersaving" mode during inactive times (night) which is why we can't re-produce the problem during the day. Perhaps you could run a couple of `amcheck's before starting amdump? That's what we do locally, because the Amanda binaries and home filesystems are NFS-auto-mounted. -- Alexandre Oliva Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/ Red Hat GCC Developer aoliva@{cygnus.com, redhat.com} CS PhD student at IC-Unicampoliva@{lsd.ic.unicamp.br, gnu.org} Free Software Evangelist*Please* write to mailing lists, not to me
Re: Failure on W2k client
Ummm! added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 read_socket_with_timeout: timeout read. read error = Connection reset by peer. session setup failed: code 0 That's Samba doing a broadcast to find a particular machine. I've noticed that amanda-2.4.2p1 doesn't like broadcasts at all and will fail dismally even though smbclient will cope. You either need to: * setup a "wins" server on the samba machine * setup /etc/smb.conf on the samba machine to point at your "wins" server Now, I can't explain this very well but Samba, by default, will lookup DNS, wins and then broadcast. You use "wins" if you want the SMB protocol to work across a subnet; because you can't broadcast across a subnet you need a special server to harvest the other subnet's information. If your local Samba can't find the machine you want locally, or via a local wins server, it will start to broadcast. Eventually one of your wins servers will respond or it will timeout. I've noticed that Amanda doesn't like this broadcast behaviour at all and will fail despite the fact that smbclient '\\OTHERMACHINE\Share' will eventually work. In my smb.conf I have: wins server = winsserver.mydomain.com.au wins support = No name resolve order = wins bcast host lmhosts Amongst other things DSL -- There's a sad face in the mirror And I'm sad to say it's me Like a ghost up in the attic Only love can set met free...
Re: Failure on W2k client
... I've noticed that amanda-2.4.2p1 doesn't like broadcasts at all and will fail dismally even though smbclient will cope. ... Huh? How can Amanda fail and smbclient work when Amanda is doing nothing more than call smbclient? What, exactly, does "fail dismally" mean, i.e. what kinds of errors are you seeing? DSL John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: Failure on W2k client
The error I see is: "Host is down or invalid password" Essentially my guess is that Amanda via smbclient attempts to use a local wins server to resolve the Netbios name which didn't work because I didn't have a wins server configured. Hence, rather than wait for the broadcast to function (it can take up to 45 seconds on my systems to do so) it just decides the host is down. When running on broadcast, a packet trace will show at least one negative smb response before manages to resolve the machine I'm backing up. Huh? How can Amanda fail and smbclient work when Amanda is doing nothing more than call smbclient? See above. DSL -- There's a sad face in the mirror And I'm sad to say it's me Like a ghost up in the attic Only love can set met free...
Re: Failure on W2k client
David Lloyd wrote: The error I see is: "Host is down or invalid password" In some sense, that is the final answer. More explicitly (and has been stated in the FAQ, and on this mailing list inumerable times): if smbclient can't get to the windows host, Amanda is not involved. You must first make smbclient work before you can even begin to test with Amanda. Again: THIS IS NOT AMANDA'S FAULT! IT IS SMBCLIENT'S FAULT! Sorry folks. I'm just really sick tired of seeing the same old problems flogged to death here because folks refuse to read either the FAQ or the archives of this mailing list. Marty P.S. Technically, it is Microsoft's fault for violating their own (unpublished) standards with the SMB implementation for w2k. P.P.S. The folks who provide Samba do an amazing job in spite of Microsoft, and if you get the very latest version from them, I'd be willing to bet a buck that it will solve your problem -- unless you have your Samba misconfigured, that is. -- Marty Shannon, RHCE, Independent Computing Consultant mailto:[EMAIL PROTECTED]
Re: Failure on W2k client
Marty! I suggest you put your asbestos, flame resistant suit on... Again: THIS IS NOT AMANDA'S FAULT! IT IS SMBCLIENT'S FAULT! You obviously cannot understand what I am saying, or you have not followed this thread. If you cannot understand English I will have this translated into any other language that I have a character set for: SMBCLIENT CAN CONTACT MY WINDOWS 2000 SERVER. IT DOES NOT FAIL. IT WORKS VIA BROADCAST AND AMANDA FAILS TO SEE THAT IT WORKS. Or to reword this: SAMBA IS ABLE TO CONTACT MY WINDOWS 2000 SERVER; SMBCLIENT CAN CONTACT IT. HOWEVER, IT DOES THIS VIA BROADCAST AND AMANDA FAILS. As far as I'm concerned SMB is working correctly. It is setup to use wins, wins fails and it therefore contacts my Windows 2000 server by broadcast and connects. Perfectly. So how is it smbclient's fault when smbclient works? I can't understand your English... DSL
Failure on W2k client
Hi Folks, Backing up shared folders (not full disk) a PC (Dell 420 dual 1ghz/W2k/sp1/15k-rpm UW scsi), we are getting this intermittent error: herman//rodney/workII lev 0 FAILED [disk //rodney/workII offline on herman?] "herman" is the amanda system ( FreeBSD 4.2-STABLE/"Amanda-2.4.2"/smbclient 2.0.7 ) this seems to be initiated by an error in sendsize (from sendsize.debug) sendsize: argument list: "smbclient" "\\rodney\workII" "-d" "0"\ "-U" "backup_user%secret" "-E" "-W" "PC-GROUP" "-c" "archive 0;recurse;du" added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 read_socket_with_timeout: timeout read. read error = Connection reset by peer. session setup failed: code 0 The multiple "added interface" messages we don't see in any successful transactions. . Repeatedly running smbclient by hand: `smbclient '\\rodney\workII' -U backup_user%secret -d 0 -E -W DOMAIN -c 'archive 0;recurse;du' we see only single interface messages, no errors or timeouts. We do notice that the response is instantaneous. Is it possible that the quick response is confusing sendsize? We are backing up other NT and W2k clients successfully and only see this error on this client. any help appreciated. thanks, hurf Hurf Sheldon Dir. Research Systems Program of Computer Graphics 580 Rhodes Hall, Hoy Rd. Cornell University Ithaca, N.Y. 14853 voice:607 255 6713 fax:607 255 0806 email: [EMAIL PROTECTED] http://www.graphics.cornell.edu/~hurf/
Re: Failure on W2k client
As someone with historic but aging knowledge of the installation in question, I'll add a couple of observations from the "for what it's worth" dept... The first thing that raises my eyebrows about this is that the address 128.84.247.9 belongs to an ethernet switch. I don't know if you changed it before posting in order to obfuscate the details for your own protection, or if that's really what was in the log file, but if it's really what's in the log file I would want to track down how amanda came up with this address. It can't be right. Secondly you said the amanda server is named "herman", but you don't have herman registered in your DNS. You didn't say who the Samba intermediary is but I'm jumping to the conclusion that you're using "herman" for that function, too. Same comment as above, perhaps you changed the name in your message to protect the innocent. I imagine the Samba intermediary machine is going to have to exist in the DNS in order for Samba backups to work properly. -Mitch On Mon, 2 Apr 2001, John R. Jackson wrote: we are getting this intermittent error: ... added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 added interface ip=128.84.247.9 bcast=128.84.247.255 nmask=255.255.255.0 read_socket_with_timeout: timeout read. read error = Connection reset by peer. session setup failed: code 0 The multiple "added interface" messages we don't see in any successful transactions. . Repeatedly running smbclient by hand: ... we see only single interface messages, no errors or timeouts. We do notice that the response is instantaneous. Is it possible that the quick response is confusing sendsize? I doubt it. The only thing I can guess is that at the time Amanda runs, something else was going on with that machine (or its network connection, etc) and Samba really couldn't talk to it. If you had tried your test right then, I'll bet it would have acted exactly like what Amanda reported. I didn't look at the Samba code, but as a guess, the multiple "added interface" lines are retries by them while trying to set up the connection, and they eventually gave up. We are backing up other NT and W2k clients successfully and only see this error on this client. That would also seem to imply something odd about this client. You might ask on the Samba mailing lists. I doubt it's a problem with their code, but they might have a better idea of what it means and if there are possible workarounds (longer timeouts, more retries, etc). hurf John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]