zfs discuss - Feb 2010 - cannot receive new filesystem stream: invalid backup stream

Can anyone suggest how I can get around the above error when 
sending/receiving a ZFS filesystem? It seems to fail when about 2/3rds 
of the data have been passed from send to recv. Is it possible to get 
more diagnostics out?

This filesystem has failed in this way for a long time, and I''ve
ignored
it thinking something might get fixed in the future, but this hasn''t 
happened yet. It''s a home directory which has been in existence and
used
for about 3 years. One thing is that the pool version (3) and zfs 
version (1) are old - could that be the problem? The sending system is 
currently running build 125 and receiving system something approximating 
to 133, but I''ve had the same problem with this filesystem for all 
builds I''ve used over the last 2 years.

-- 
Cheers
Andrew Gabriel

On 12/02/2010 09:55, Andrew Gabriel wrote:
> Can anyone suggest how I can get around the above error when
> sending/receiving a ZFS filesystem? It seems to fail when about 2/3rds
> of the data have been passed from send to recv. Is it possible to get
> more diagnostics out?
You could try using /usr/bin/zstreamdump on recent builds.
> This filesystem has failed in this way for a long time, and I''ve
ignored
> it thinking something might get fixed in the future, but this
hasn''t
> happened yet. It''s a home directory which has been in existence
and used
> for about 3 years. One thing is that the pool version (3) and zfs
> version (1) are old - could that be the problem? The sending system is
> currently running build 125 and receiving system something approximating
> to 133, but I''ve had the same problem with this filesystem for all
> builds I''ve used over the last 2 years.
>

-- 
Darren J Moffat

Darren J Moffat wrote:
> On 12/02/2010 09:55, Andrew Gabriel wrote:
>> Can anyone suggest how I can get around the above error when
>> sending/receiving a ZFS filesystem? It seems to fail when about 2/3rds
>> of the data have been passed from send to recv. Is it possible to get
>> more diagnostics out?
>
> You could try using /usr/bin/zstreamdump on recent builds.
Ah, thanks Darren, didn''t know about that.

As far as I can see, it runs without a problem right to the end of the 
stream. Not sure what I should be looking for in the way of errors, but 
there are no occurrences in the output of any of the error strings 
revealed by running strings(1) on zstreamdump. Without -v ...

zfs send export/home at 20100211  |  zstreamdump                      
BEGIN record
    version = 1
    magic = 2f5bacbac
    creation_time = 4b7499c5
    type = 2
    flags = 0x0
    toguid = 435481b4a8c20fbd
    fromguid = 0
    toname = export/home at 20100211
END checksum = 
6797d43d709150c1/e1ec581cbaf0cfa4/7f6d80fa0f23c741/2c2cb821b4a2e639
SUMMARY:
    Total DRR_BEGIN records = 1
    Total DRR_END records = 1
    Total DRR_OBJECT records = 90963
    Total DRR_FREEOBJECTS records = 23389
    Total DRR_WRITE records = 212381
    Total DRR_FREE records = 520856
    Total records = 847591
    Total write size = 17395359232 (0x40cd81e00)
    Total stream length = 17683854968 (0x41e0a3678)
>> This filesystem has failed in this way for a long time, and
I''ve ignored
>> it thinking something might get fixed in the future, but this
hasn''t
>> happened yet. It''s a home directory which has been in
existence and used
>> for about 3 years. One thing is that the pool version (3) and zfs
>> version (1) are old - could that be the problem? The sending system is
>> currently running build 125 and receiving system something
approximating
>> to 133, but I''ve had the same problem with this filesystem for
all
>> builds I''ve used over the last 2 years.
-- 
Andrew