Conference decwet::advfs_support

Hello, thanks for your note.

Are you able to reproduce the problem as stated?  Also, does the problem occur

when you use init 0 to shutdown the second system (having relocated the 
service to the first)?

Sometimes the "bs_osf_complete: metadata write failed" error originates
from a hardware problem.  Do you show any such errors in the uerf logs?

Thanks,
Janice




[Posted by WWW Notes gateway]

Hi Janice,

Thanks for reply.

I tryed on office : init 0 to halt system  give metadata write failed
                    shutdown .... no errors ...

I have the deamon.lod and uerf for the two customer system but
can reproduce at office the customer problem.... that doesn't
look to hardware problem because I checked all on office and no hardware
errors.

Summary of my test on Office:

sys 1      sys 2          Command used on sys2   Result
                            to halt system
=============================================================================
NFSservice                on sys2 :shutdown      No metadata write failed

NFSservice                on sys2 :init 0        METADATA WRITE FAILED ...
<<---

           NFSservice     on sys2 :shutdown      No metadata write failed

           NFSservice     on sys2 :init 0        No metadata write failed


In this way, I found (and I can reproduce) than METADATA WRITE FAILED 
is displayed  ONLY if ase service is on the other member and only if you have
used init 0 for halt system ....

Hope that may help you ...

Thierry


[Posted by WWW Notes gateway]

Thierry,

I've been discussing this problem with a few other people.  One of them
suggested it may be a problem with HSZs timing out devices too late for ASEs
convenience.

You say you can reproduce the problem on your systems.  Do they include HSZ40s
or HSZ50s?

Do you have any accompanying SCSI_STAT_RESERVATION_CONFLICT messages in your
system error log?

Herta

Herta,

If I try to do a uerf -R -o full |more ... I get ...

----- EVENT INFORMATION -----

EVENT CLASS                             ERROR EVENT
OS EVENT TYPE                  199.     CAM SCSI
SEQUENCE NUMBER                 10.
OPERATING SYSTEM                        DEC OSF/1
OCCURRED/LOGGED ON                      Tue Mar 11 15:47:07 1997
OCCURRED ON SYSTEM                      tscosf
SYSTEM ID                 x00020004     CPU TYPE:  DEC 3000
SYSTYPE                   x00000000

----- UNIT INFORMATION -----

CLASS                         x0000     DISK
SUBSYSTEM                     x0000     DISK
BUS #                         x0002
                              x00A8     LUN x0
                                        TARGET x5

----- CAM STRING -----

ROUTINE NAME                            cdisk_op_spin

----- CAM STRING -----

                                        Unit Reserved

----- CAM STRING -----

ERROR TYPE                              Information Message Detected
                                         _(recovered)

----- CAM STRING -----

DEVICE NAME                             DEC     RZ28B

----- CAM STRING -----

                                        Active CCB at time of error

----- CAM STRING -----

                                        CCB request completed with an error
ERROR - os_std, os_type = 11, std_type = 10


----- ENT_CCB_SCSIIO -----

*MY ADDR                  x09F82F28
CCB LENGTH                    x00C0
FUNC CODE            x01
CAM_STATUS                    x0004     CAM_REQ_CMP_ERR
PATH ID              2.
TARGET ID            5.
TARGET LUN           0.
CAM FLAGS                 x000004C0
                                        CAM_DIR_NONE
                                        CAM_SIM_QFRZDIS
*PDRV_PTR                 x09F82C28
*NEXT_CCB                 x00000000
*REQ_MAP                  x00000000
VOID (*CAM_CBFCNP)()      x00521BB0
*DATA_PTR                 x00000000
DXFER_LEN                 x00000000
*SENSE_PTR                x09F82C50
SENSE_LEN            x40
CDB_LEN              x06
SGLIST_CNT                    x0000
CAM_SCSI_STATUS               x0018     SCSI_STAT_RESERVATION_CONFLICT
SENSE_RESID          x00
RESID                     x00000000
CAM_CDB_IO           x000000000000000000000000
CAM_TIMEOUT               x00000014
MSGB_LEN                      x0000
VU_FLAGS                      x0000
TAG_ACTION           x00

///Thierry

Janice,

I called Herta .. and the difference between Herta problem and I is only
that Herta is using KZP.. 
But we have same problem ... Metadata write failed if service is on second
member and that we are using init 0 on first system. After that,
a look to binary.errlog file showed us SCSI_STAT_RESERVATION_CONFLICT 

///Thierry




[Posted by WWW Notes gateway]

Hello  Thierry.  I have not been able to find previous cases dealing
with a similar problem.  I'll have to recommend that an IMPT case be
filed, since it doesn't look like there is a known solution.

Thanks,
Janice

[Posted by WWW Notes gateway]

    Thiery,
    
    PLEASE obtain the crash-data file and send it to the CANASTA Mail
    Server with the following command:
    
    # Mail -s "Diagnose case=advfs_1015 customer=notes_on_decwet" 
    			can_server@xocomp.enet.dec.com < crash-data.n
    
    This will run your crash-data file through CANASTA and - even if no
    solution might be available - will make SURE, that this FOOTPRINT is
    entered into the CANASTA case database.
    
    Thanks,
    
    Volker.
    
    PS: Please read note TURRIS::DIGITAL_UNIX 8919 for more info about CANASTA

Hi,

I'm %%% NOT %%% able to send crash data because there is %%%NO%%%
crash-data file ... and no vmcore and vmunix ...

In fact, when system is shutting down, he display on screen

advfs I/O error: setId 0x33120b87.0002edd5.fffffffe.0000  tag 0xfffffff7.0000u
    page 42
            vd 1  blk 2944  blkCnt 16
            write error = 5

    bs_osf_complete: metadata write failed
    AdvFS Domain Panic; Domain nmcs-dom1 Id 0x33120b87.0002edd5

more than 2 or 3 times but there are NOTHING in /var/adm/crash.

The only one thing that I can give you is the binary.errlog file.
In these file, you may see the SCSI_STAT_RESERVATION_CONFLICT.

I spoke with one of my collegas and he said me that problem is
maybe in relation with the order of K... scripts in /sbin/rc0.d

Thanks to all,
Thierry

[Posted by WWW Notes gateway]