| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 319.1 | Before you IPMT, please give us (much) more information | STOWKS::SLUIS | Hans van Sluis -- Storage Engineering Support Europe- DTN 889 9526 | Mon Feb 17 1997 05:04 | 21 |
| John,
This *CAN* be normal if the disk becomes full and HSM has to move files to the
cache or (worse) to tape.
Therefore, can you do a '$SHOW DEVICE D' of your user disks.
Furthermore, did you provide the output of the following :
SMU SHOW FACILITY
SMU SHOW VOLUME
SMU SHOW POLICY
SMU SHOW SHELF
SMU SHOW CACHE
SMU SHOW ARCHIVE
SMU SHOW DEVICE
You also should have the log-files in HSM$LOG ready for inspection.
Please post the output over here, so that we have a better impression of
what we're talking about.
Hans van Sluis, StorageWorks Engineering Support Europe
|
| 319.2 | details | COMEUP::SIMMONDS | lock (M); while (not *SOMETHING) { Wait(C,M); } unlock(M) | Mon Feb 17 1997 20:01 | 125 |
| Hans, here's the stuff you asked for.. the disk volume in question is
DRB0: .. the HSM cache is on DRB1: .. at no time do _any disks ever_
exceed 75% used space!! .. the Customer is typically running a DCL
procedure which runs a program which causes a file fault on a
particular file (using sys$crmpsc()) in a DRB0: directory and then
performs an explicit $ SHELVE <file> after the program has exited.
John.
----<i>---------------------------------------------------------------
$smu show facility
Polycenter HSM is enabled for Shelving and Unshelving
Facility history:
Created: 10-FEB-1997 12:22:25.42
Revised: 14-FEB-1997 20:34:59.18
Designated servers: Any cluster member
Current server: ARCHER
Catalog server: Enabled
Event logging: Audit
Error
Exception
HSM mode: Basic
Remaining license: 20 gigabytes
$smu show volume
Volume HSM$DEFAULT_VOLUME on Shelf HSM$DEFAULT_SHELF, Shelving is enabled,
Unshelving is enabled, Highwater mark detection is disabled, Occupancy full
detection is disabled, Disk quota exceeded detection is disabled
Volume _ARCHER$DRB0: on Shelf CRISP$ARCHIVE, Shelving is enabled, Unshelving is
enabled, Highwater mark detection is disabled, Occupancy full detection
is disabled, Disk quota exceeded detection is disabled
$smu show policy
Policy HSM$DEFAULT_OCCUPANCY is enabled for shelving
Policy HSM$DEFAULT_POLICY is enabled for shelving
Policy HSM$DEFAULT_QUOTA is enabled for shelving
$smu show shelf
Shelf CRISP$ARCHIVE is enabled for Shelving and Unshelving
Catalog File: DISK$DATA:[HSM$SERVER.CATALOG]HSM$CATALOG.SYS
Shelf History:
Created: 10-FEB-1997 12:22:25.22
Revised: 10-FEB-1997 12:22:25.22
Backup Verification: Off
Save Time: <none>
Updates Saved: All
Archive Classes:
Archive list: HSM$ARCHIVE01 id: 1
Restore list: HSM$ARCHIVE01 id: 1
Shelf HSM$DEFAULT_SHELF is enabled for Shelving and Unshelving
Catalog File: DISK$DATA:[HSM$SERVER.CATALOG]HSM$CATALOG.SYS
Shelf History:
Created: 10-FEB-1997 12:22:25.41
Revised: 10-FEB-1997 12:22:25.41
Backup Verification: Off
Save Time: <none>
Updates Saved: All
Archive Classes:
Archive list: <none>
Restore list: <none>
$smu show cache
Cache device _ARCHER$DRB1: is enabled, Cache flush is held until after
15-JAN-1997 15:40:10.35, Backup is performed at flush intervals,
Cached files are held on delete of online file
Block size: 500000
Highwater mark: 80%
Flush interval: 0 01:00:00.00
$smu show archive
HSM$ARCHIVE01 has been used
Identifier: 1
Media type: CompacTape III
Label: HSM002
Position: 96
Device refs: 1
Shelf refs: 2
HSM$ARCHIVE02 has been used
Identifier: 2
Media type: CompacTape III
Label: HSM002
Position: 99
Device refs: 1
Shelf refs: 0
$smu show device
HSM drive HSM$DEFAULT_DEVICE is enabled.
Shared access: < shelve, unshelve >
Drive status: Not configured
Media type: Unknown Type
Robot name: <none>
Enabled archives: <none>
HSM drive _ARCHER$MKA200: is enabled.
Shared access: < shelve, unshelve >
Drive status: Configured
Media type: CompacTape III
Robot name: <none>
Enabled archives: HSM$ARCHIVE01 id: 1
HSM$ARCHIVE02 id: 2
$show device d
Device Device Error Volume Free Trans Mnt
Name Status Count Label Blocks Count Cnt
ARCHER$DKB500: Online wrtlck 0
ARCHER$DRA0: Mounted 0 OPENVMS 4978287 405 1
ARCHER$DRA1: Mounted 0 TRAN 6171021 2 1
ARCHER$DRA2: Mounted 0 SECS 7325100 1 1
ARCHER$DRB0: Mounted 0 DATA 2562417 11 1
ARCHER$DRB1: Mounted 0 DEV 5564898 9 1
ARCHER$DVA0: Online 0
|
| 319.3 | IPMT launched | COMEUP::SIMMONDS | lock (M); while (not *SOMETHING) { Wait(C,M); } unlock(M) | Mon Feb 24 1997 19:54 | 36 |
| Ok, IPMT on it's way.
The observed stalls are totally unacceptable for Production or even
Development deployment of HSM.
To summarize:
[ AlphaServer 1000A 5/300 ]=========== [ KZPSA ]
| |
"OPENVMS" DRA0 ---| |--- DRB0 "DATA" <----+
| | :
"TRAN" DRA1 ---| |--- DRB1 "DEV" <-+ :
| + : :
"SECS" DRA2 ---| : :
+ : :
: :
HSM Cache dev. --+ :
:
HSM Shelv. Vol ------+
1. User runs a batch procedure which first runs a user image which
file faults a single file on DRB0: (via sys$crmpsc()); file size
is 495000 blocks.
2. After doing some comptations on the file contents, the image exits.
3. The DCL batch procedure next explicitly SHELVEs the file that was
just used, selects a new (shelved) file to use and loops to (1.)
These are the _only_ user-initiated HSM operations occurring on DRB0:
Now, if someone happens to reference some file on DRB0: (_not_ the files
touched by the batch procedure!) then, *sometimes* all I/O on that disk
is totally stalled for about 160..170 seconds.
John.
|
| 319.4 | -- this is 319.0 -- | COMEUP::SIMMONDS | lock (M); while (not *SOMETHING) { Wait(C,M); } unlock(M) | Mon Feb 24 1997 19:57 | 16 |
| Before sending this into IPMT-land..
On a shared volume (i.e. general user disk) where HSM Unshelving/Shelving
activity is occurring in one directory, _ALL_ user I/O to the volume
occasionally gets stalled for 160..170 *Seconds* ? Surely this is not
normal ?
[Environment:
This is the same node reported in note 318.0 --
AlphaServer 1000A, 512Mb, SWXCR -> 5 JBOD drives 4Gb each
OpenVMS Alpha Version V6.2-1H3
HSM V2.0A (MUP)]
Thanks!
John.
|
| 319.5 | | COMEUP::SIMMONDS | lock (M); while (not *SOMETHING) { Wait(C,M); } unlock(M) | Mon Mar 10 1997 02:04 | 18 |
| Ooops.. (*blush*)
The .3 configuration is wrong; here is the correct disk config.:
[KA1B05]====PCI==============[KZPSC #0]==========[KZPSC #1]========...
| |
| |
|-DRA0 |-DRB0
|-DRA1 |-DRB1
|-DRA2
(All DR disks are single JBOD, i.e. No RAID)
Also, (from a proposed explanation in an IPMT Update mail message)
the behaviour may be due to file highwater marking enabled for the HSM
shelving volume.. testing in progress.. will reply with more news.
John.
|
| 319.6 | probably FHWM.. | COMEUP::SIMMONDS | lock (M); while (not *SOMETHING) { Wait(C,M); } unlock(M) | Mon Mar 10 1997 03:11 | 7 |
| Ok, the stalls have gone since we turned file highwater marking OFF !
Thanks for your forbearance.. I'll check with the Customer whether they
really _require_ File HWM on this volume in their proposed Production
environment before I make any further comment..
John.
|