When executing the command devbra -dev on Windows 2008 you may get no results displayed. In this case the user account control is the source of the problem. Open the command prompt as administrator and you will get the resuklt displayed.
Probably most customers using Data Protector with a Fibre Channel Tape Library attached expierienced this problem: a SCSI Reset during the backup including destroyed media. What happened? The drive was flooded with SCSI commands – device polling was the cause.
Polling is the behaviour of a service which polls tape drives and library controller via IFC (Interface Controller) or NSR (Network Storage Router) using SCSI commands frequently. This may influence running backups, as the NSR must process additional commands, which mostly come not from the backup software.
- Changed backup speed
- I/O errors
- The backup software recognizes the media as poor
- A tape drive will be set to offline
- The transfer rated decreases
- The tape drive can not stream
Known serives causing these errors:
- Network scanner
- Removable Storage Manager
- Tape driver
- OVOSAM using old Tachyon drivers
- Insight Management Agents (SIM)
- Retrosopect / Retrorun (Macintosh)
Recommendation: Deactivate polling on all servers where tape drives are attached.
In the article “What is device polling” I described what device polling is and what can happen to your backup when polling occurs. I now describe how device polling can be prevented or eliminated.
The following is valid for HP hardware only. If you need to use Data Protector on F*****, I** or D*** servers, please refer to the documentation of your hardware verndor.
As a common rule deactivate device polling on all servers sharing the SAN with library or tape devices. Servers which must not see tape devices or libraries should be removed from zoning used by your tape devices and libraries. If possible use 1:1 zoning. For MSL 6000 libraries create an empty indexed map to prevent new servers to automatically zoned to your devices. Create a separate map on your NSR.
- Deactivate network scanners: NMAP or 3com Network Director is known to send data to the NSR, which may lead to a reboot of the NSR. Do scans after the backup is done.
- In Windows the so called TUR commands are sent on the SCSI bus periodically (Windows 2000 – every 3 seconds, Windows 2003, every second). This behaviour starts immediatley when you install a new server with SAN attached. This applies also for Windows 2008 when the removable stroages service is installed as an additional feature (not default). As a general advise deactivate removable stroages service and disable TUR on tape devices (see below). Microsoft release article KB842411: http://support.microsoft.com/default.aspx?scid=kb;en-us;842411. Additional hint: If you deactivate a server to be used for new applications, make sure you remove this server from zoning as polling will occur immediately when you install the server.
- For HP LTO drives install the latest driver version as an old version (22.214.171.124) has polling activated within the install file. In general the following Regkey must be created for each drive you have (drive type – hplto, hpdlt, …), when donw restart the server. Sometime it might be necessary to reboot the NSR as well. The steps must also be done for the robotic control (libxprmc, hpmc or hpuln32).
- Create new DWORD, AutoRun, Value=0
- Insight Management Agents can send requests to tape devices to recognize errors with your tapes. In control panel open HP Management Agents, navigate to tab Storage and select “Disable Fibre Agent Tape Support”.
On Unix device polling is also a problem:
- For HP-UX 11.23 (v2) you have to set the kernel parameter st_san_safe –>
- On HP-UX 11.31 (v3) there is a new parameter to prevent device polling, patch PHKL_36312 is a requirement:
- Change the parameter:
scsimgr set_attr -d estape -a norewind_close_disabled=1
- Save the parameter:
scsimgr save_attr -d estape -a norewind_close_disabled=1
- Check the parameter:
scsimgr - estape get_attr
- Change the parameter:
And what can you do in addition? Empty the traces log on the NSR and check how many entries are written when no backup is running. When device polling occurs you will notice a lot of messages per second here.
Comment: Polling will still happen, but not with the problems described above. When running “ioscan” on HP-UX or “search for changed hardware” on Windows or when running LTT, you might still see some device polling, but this is a normal level.
And last comment as I talked about it… LTT (when used in current version) can document which HBA did TUR, with a good SAN documentation it is very easy tio identify the host which does polling on your devices. LTT can be downloaded here: (Did you know that LTT is available for Netware? – I’m a Master CNE…)http://h20000.www2.hp.com/bizsupport/TechSupport/DriverDownload.jsp?lang=de&cc=de&prodNameId=406731&taskId=135&prodTypeId=12169&prodSeriesId=406729&lang=de&cc=de
With document ID c02618631 on 2010/11/19 an advisory for HP Data Protector 6.11 was released, which describes a problem when using the GUI after installing VEPA patch DPWIN_00483; the creation of bar backups is no longer possible, as browsing of instances is not possible.
There is a solution, copying back the saved GUI parts and installing the VEPA patch to dedicated clients only.
The complete advisory can be read here: http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=c02618631&dimid=1004694894&dicid=alr_nov10&jumpid=em_alerts/us/nov10/all/xbu/emailsubid/mrm/mcc/loc/rbu_category/alerts
UPDATE: DPWIN_490 solves the problem, please refer to “patches”.
Once you installed the patch DPWIN_00483 (VEPA patch) you are no longer able to create new integration backups (Exchange, SQL, …). Withhin GUI you are not able to browse for instances. The bug comes from changed manager.exe through patch DPWIN_00483. As a workaround you may use the JAVA GUI where the bug was not seen so far. HP also released an advisory – see corresponding section.