--On Saturday, July 30, 2005 15:09:26 -0400 "Scott R. Burns" <Scott.Burns AT
Netcontech DOT Com> wrote:
> Thanks. I was under the assumption that it would just switch the tapes and
> restart that dump from the beginning on the new tape. Reading through the
> code it kind of looks like it performs all "direct to tape" dumps after the
> holding disk dumps and I was wondering if those were retried. I am not 100%
> sure on this yet.
I haven't poked through the code much, so I'll let someone else answer on how
it actually decides on when the direct to tape dumps are done.
>
> Can I assume that dumporder is related to disklist order then ? Or do I need
> to also mark it's priority high for example ?
Dump order is determined by the 'dumporder' variable in your config file. I
don't
recall what it's default value is. There is also a 'taperalgo' variable that
can
be set to 'largestfit' that may help in your case. However, if your code
interpretation
is correct and all holdingdisk DLEs are done first, and if that one direct to
tape DLE
is all that's remaining, and you haven't yet written 'length' bytes to the
tape, I'm
not sure if Amanda will go ahead and change tapes since the first one isn't
full.
Perhaps one of the folks more involved in code development can elaborate on
the inner workings of how all these things interact when 'direct to tape' dumps
are involved. All my experience is with everything going to the holdingdisk
first.
Frank
>
> Thanks for your reply.
>
> Scott...
>
> Scott R. Burns
> NETCON Technologies Inc.
> Suite 135 - 4474 Blakie Road
> London, Ontario, Canada
> N6L 1G7
> Voice: +1.519.652.0401
> Fax: +1.519.652.9275
>
>
> -----Original Message-----
> From: Frank Smith [mailto:fsmith AT hoovers DOT com]
> Sent: Saturday, July 30, 2005 2:54 PM
> To: Scott R. Burns; amanda-users AT amanda DOT org
> Subject: Re: Problem with runs requiring multiple tapes
>
>
> --On Saturday, July 30, 2005 10:39:52 -0400 "Scott R. Burns"
> <Scott.Burns AT Netcontech DOT Com> wrote:
>
>> I am running on:
>>
>> NetBSD V2.0.2/i386
>> HP Surestore DAT24x6
>> Amanda-2.4.4p4
>>
>> I have setup a daily backup config which runs well, but it all fits on one
>> tape.
>>
>> I have setup parallel monthly backup config that forces level 0s and uses
> no
>> record and this does not seem to be working properly. This config has to
> go
>> to a 2nd tape and never does. I have tested the drive and tapes and all
> work
>> ok. All paritions dump properly until the last one. On the last parition
> it
>> tries, which compresses larger than my holding area, but smaller than a
>> tape, here is what I see in the logs around the time of the failure:
>>
>
> I think you need to either change your dumporder so that your largest DLE
> (this
> failing one) goes first so it can all be written on the first tape, increase
> your
> holdingdisk space so that it will fit, or split this DLE into pieces that
> will fit
> on your holdingdisk.
> Currently, it is dumping directly to tape since it is larger than your
> holdingdisk
> but it is hitting EOT. If it were on the holdingdisk, Amanda could just
> start over
> on the next tape, but since it is dumping directly to tape it can't, and
> just fails
> that DLE. I don't think Amanda has a way to tell the client to abort the
> currently
> running dump and start over.
>
> Frank
>
>
>> amdump.1
>> ---------
>>
>> (successful lines trimmed...here is where the issue starts)
>> driver: dumping ncti103:sd1f directly to tape
>> driver: send-cmd time 7660.547 to taper: PORT-WRITE 00-00035 ncti103
>> fffffeff9ffe0f sd1f 0 20050729
>> taper: try_socksize: receive buffer size is 65536
>> taper: stream_server: waiting for connection: 0.0.0.0.61737
>> driver: result time 7660.548 from taper: PORT 61737
>> driver: send-cmd time 7660.548 to dumper0: PORT-DUMP 01-00036 61737
> ncti103
>> fffffeff9ffe0f sd1f NODEVICE 0 1970:1:1:0:0:0 DUMP
>>| ;bsd-auth;compress-fast;no-record;index;
>> taper: stream_accept: connection from 127.0.0.1.61736
>> taper: try_socksize: receive buffer size is 32768
>> driver: state time 7660.548 free kps: 4370 space: 7840470 taper: writing
>> idle-dumpers: 7 qlen tapeq: 0 runq: 0 roomq: 0 wakeup: 86400 driver-idle:
>> not-idle
>> driver: interface-state time 7660.548 if : free 970 if FXP1: free 1000 if
>> BGE0: free 1000 if LE0: free 400 if LOCAL: free 1000
>> driver: hdisk-state time 7660.548 hdisk 0: free 7840470 dumpers 0
>> dumper: stream_client: connected to 127.0.0.1.61737
>> dumper: stream_client: our side is 0.0.0.0.61736
>> dumper: try_socksize: send buffer size is 65536
>> dumper: stream_client: connected to 200.1.1.103.61735
>> dumper: stream_client: our side is 0.0.0.0.61732
>> dumper: stream_client: connected to 200.1.1.103.61734
>> dumper: stream_client: our side is 0.0.0.0.61731
>> dumper: stream_client: connected to 200.1.1.103.61733
>> dumper: stream_client: our side is 0.0.0.0.61730
>> taper: writing end marker. [nctimonthly000 ERR kb 11990208 fm 18]
>>
>> (Above: appears to have run out of tape on the first tape here. What does
>> ERR mean ?)
>>
>> changer: opening pipe to: /usr/pkg/libexec/chg-scsi -info
>> changer: got exit: 0 str: 0 5 1 0
>> changer: opening pipe to: /usr/pkg/libexec/chg-scsi -slot current
>> changer: got exit: 0 str: 0 /dev/nrst0
>> taper: slot 0: date 20050729 label nctimonthly000 (active tape)
>> changer: opening pipe to: /usr/pkg/libexec/chg-scsi -slot next
>> changer: got exit: 0 str: 1 /dev/nrst0
>> taper: slot 1: date X label nctimonthly001 (new tape)
>> taper: read label `nctimonthly001' date `X'
>> taper: wrote label `nctimonthly001' date `20050729'
>>
>> (Above: now it has found the next tape, writes the label. Never seems to
>> retry after this.)
>>
>> driver: result time 12648.338 from dumper0: FAILED 01-00036 ["data write:
>> Broken pipe"]
>> driver: result time 12648.338 from taper: TRY-AGAIN 00-00035 [writing
> file:
>> Input/output error]
>>
>> (Above: is this indicating the 2nd tape failed, or just logging the first
>> failure ?)
>>
>> dumper: kill index command
>> driver: state time 12648.383 free kps: 4400 space: 7840470 taper: idle
>> idle-dumpers: 8 qlen tapeq: 0 runq: 0 roomq: 0 wakeup: 86400 driver-idle:
>> not-idle
>> driver: interface-state time 12648.383 if : free 1000 if FXP1: free 1000
> if
>> BGE0: free 1000 if LE0: free 400 if LOCAL: free 1000
>> driver: hdisk-state time 12648.383 hdisk 0: free 7840470 dumpers 0
>> driver: QUITTING time 12648.383 telling children to quit
>> driver: send-cmd time 12648.383 to dumper0: QUIT
>> driver: send-cmd time 12648.387 to dumper1: QUIT
>> driver: send-cmd time 12648.388 to dumper2: QUIT
>> driver: send-cmd time 12648.388 to dumper3: QUIT
>> driver: send-cmd time 12648.388 to dumper4: QUIT
>> driver: send-cmd time 12648.389 to dumper5: QUIT
>> driver: send-cmd time 12648.389 to dumper6: QUIT
>> driver: send-cmd time 12648.390 to dumper7: QUIT
>> driver: send-cmd time 12648.390 to taper: QUIT
>> taper: DONE [idle wait: 285.162 secs]
>> taper: writing end marker. [nctimonthly001 OK kb 0 fm 0]
>>
>> (Above: writes the end marker here but it never wrote any data to the tape
>> ?)
>>
>> driver: FINISHED time 12659.810
>> amdump: end at Fri Jul 29 12:56:50 EDT 2005
>>
>>
>> My first 3 attempts to post failed, hopefully this one makes it. I have
> all
>> the logs from these runs. I have tried this 5 times with the same results.
> I
>> can post them on ftp if reqired.
>>
>> Scott...
>>
>> Scott R. Burns
>> NETCON Technologies Inc.
>> Suite 135 - 4474 Blakie Road
>> London, Ontario, Canada
>> N6L 1G7
>> Voice: +1.519.652.0401
>> Fax: +1.519.652.9275
>>
>
>
>
> --
> Frank Smith
> fsmith AT hoovers DOT com
> Sr. Systems Administrator Voice:
> 512-374-4673
> Hoover's Online Fax:
> 512-374-4501
--
Frank Smith fsmith AT hoovers
DOT com
Sr. Systems Administrator Voice: 512-374-4673
Hoover's Online Fax: 512-374-4501
|
