Odd DSSI/Cluster behaviour. Phanton disks appear online
- From: JF Mezei <jfmezei.spamnot@xxxxxxxxxxxxx>
- Date: Tue, 16 Jan 2007 08:06:40 -0500
This in not important, but interesting/ odd !
Node Velo and Wheel (VAX 7.3) have shared DSSI access to 5 drives $4$dia1 to $4$dia5.
Nodes Chain and Bike are Alphas at 8.3
4 of those drives are dismounted from all 4 nodes.
The 4 drives are physically taken out of the DSSI slots.
All 4 nodes now show them as HostUnavailable.
Velo is rebooted. It, and the 2 alphas now see those 4 drives as on-online and served by the newly rebooted VELO ! Wheel still sees them as HostUnavailable.
I assume WHEEL made those drives available to VELO which ignored the HostUnavailable status and announced it could serve those drives too without checking that it could in fact access them. So Chain and Bike though the drives were on-line again !
I tried to mount one of the drives for fun, and it asked me to load the device (like for a tape).
----
Now, I shutdown both VELO and WHEEL at the same time and rebooted them (after disconnecting the 5th drive which was their system drive). Both rebooted without any knowledge of those 5 drives.
On the alphas, the drives remain seen (normal since disk drives never go away), but they are still shown as ONLINE ! None of the VAXes know about those devices since when they rebooted, these was no trace of those drives.
SYSMAN> do show dev $4$d
%SYSMAN-I-OUTPUT, command execution on node BIKE
Device Device Error Volume Free Trans Mnt
Name Status Count Label Blocks Count Cnt
$4$DIA1: (VELO) Online 0
$4$DIA2: (VELO) Online 0
$4$DIA3: (VELO) Online 0
$4$DIA4: (VELO) Online 0
$4$DIA5: (WHEEL) Online 0
%SYSMAN-I-OUTPUT, command execution on node CHAIN
Device Device Error Volume Free Trans Mnt
Name Status Count Label Blocks Count Cnt
$4$DIA1: (VELO) Online 0
$4$DIA2: (VELO) Online 0
$4$DIA3: (VELO) Online 0
$4$DIA4: (VELO) Online 0
$4$DIA5: (WHEEL) Online 0
%SYSMAN-I-OUTPUT, command execution on node VELO
%SYSTEM-W-NOSUCHDEV, no such device available
%SYSMAN-I-OUTPUT, command execution on node WHEEL
%SYSTEM-W-NOSUCHDEV, no such device available
I guess that since neither VELO or WHEEL know about those devices, they are not sending any messages to the rest of the cluster to advise they are offline.
Now, if I try to mount it on an alpha, I get:
$ mount $4$dia1/override=id
%MOUNT-F-MEDOFL, medium is offline
Disk $4$DIA1: (VELO), device type RF73, is online, file-oriented device,
shareable, available to cluster, error logging is enabled.
Error count 0 Operations completed 3244
Owner process "" Owner UIC [SYSTEM]
Owner process ID 00000000 Dev Prot S:RWPL,O:RWPL,G:R,W
Reference count 0 Default buffer size 512
Current preferred CPU Id 0 Fastpath 1
Total blocks 3906420 Sectors per track 71
Total cylinders 2620 Tracks per cylinder 21
Host name "VELO" Host type, avail VAX 4000-600A, yes
Alternate host name "WHEEL" Alt. type, avail VAX 4000-200, yes
Allocation class 4
You'd think that after a failed mounting attempt, the device would be marked as "offline" and host-unavailable.
.
- Follow-Ups:
- Re: Odd DSSI/Cluster behaviour. Phanton disks appear online
- From: baldrick
- Re: Odd DSSI/Cluster behaviour. Phanton disks appear online
- From: etmsreec
- Re: Odd DSSI/Cluster behaviour. Phanton disks appear online
- Prev by Date: Re: Audio Cast #2 is now available
- Next by Date: Re: Blast from the 1988s (DEC proposal)
- Previous by thread: Re: Audio Cast #2 is now available
- Next by thread: Re: Odd DSSI/Cluster behaviour. Phanton disks appear online
- Index(es):
Relevant Pages
|