Date: 2012-06-15
OS: SLES
OS Version: 11 SP1
Symantec FileStore 
5.7 P3 Patch Upgrade README

CONTENTS
I.   Overview
II.  Upgrade procedure
III. Fixes in the new patch
IV.  Known issue
V.   New Features

PATCH ID 		        : N/A
PATCH NAME 		        : sfs-sles11_x86_64-5.7P3-patches.tar.gz
BASE PACKAGE NAME	    : Symantec FileStore
BASE PACKAGE VERSION 	: 5.7
OBSELETE PATCHES 	    : N/A
SUPERCEDED PATCHES 	    : N/A
INCOMPATIBLE PATCHES 	: N/A
SUPPORTED OS 		    : SLES
SUPPORTED OS VERSION	: SLES 11 SP1
CREATION DATE 		    : 2012-06-15
CATEGORY 		        : enhancement, performance issue
REBOOT REQUIRED 	    : Yes
SUPPORTS ROLLBACK 	    : NO


I. OVERVIEW:
------------
Symantec FileStore provides a scalable clustered storage solution. This document provides release information for the patch.

Notes:
1. Patch upgrade will cause service failover from one node to another which will take about 15 seconds. This will not impact NFS client mounts and replication jobs, but CIFS client mounts will need to connect again, and backup jobs need to be resumed as well after an upgrade. 
2. For the above reason, it is recommended to stop CIFS, backup, and EV I/O loads during an upgrade.  
3. The 5.7 P3 patch can only be applied to the 5.7, 5.7 P1, or 5.7P2 release.  
4. Patch upgrade cannot be rollbacked once applied.
5. Don't proceed upgrade when there are heavy I/O ongoing

II. UPGRADE PROCEDURE:
----------------------
To install the patch:
1. Copy SFSFS-patch-5.7P3_rc6_2012_06_13.tar.gz to the FileStore master node at (/home/support/patches/) using the support login.
2. If fencing is enabled, turn off and remove I/O fencing before the upgrade:
filestore> storage> fencing off
filestore> storage> fencing destroy

3. Install the FileStore patch using the master login (CLISH):

filestore>upgrade patch install scp://support@localhost:/home/support/patches/SFSFS-patch-5.7P3_rc6_2012_06_13.tar.gz

The patch is installed on the slave nodes first followed by a system reboot. The master node is installed last followed 
by a system reboot.

4. If fencing was enabled, enable it again.

5. Monitor and check the status of the FileStore cluster: 

To check that the installation is complete, login to the master node again (switch to another node), and check that all the nodes are in the RUNNING state:

filestore> cluster show
------------       -----      -----------   -------- --------   -------- --------
filestore_01       RUNNING          0.98       0.01     0.00       0.01     0.00 
filestore_02       RUNNING          0.97       0.01     0.00       0.01     0.00 

filestore_01# gabconfig -a
GAB Port Memberships
=================================
Port a gen   187002 membership 01
Port b gen   187004 membership 01
Port f gen   18700e membership 01
Port h gen   187006 membership 01
Port u gen   18700c membership 01
Port v gen   187008 membership 01
Port w gen   18700a membership 01

Note: Please refer to the Symantec FileStore 5.7 Command-Line Administrator's Guide for information on the "support" 
user login and more information about the patch installation.


III. FIXES IN THE NEW PATCH:
----------------------------

Etrack Incidents:  2780959, 2725972, 2776252, 2590189, 2779856, 2767960, 2744096, 2809277, 2761697, 2431150, 2780945, 2772235, 2429200, 2761702, 2763413, 2771990, 2796103, 2592820, 2740620, 2633743, 2633682, 2783403, 2783407, 2795588, 2795076, 2638340, 2637979, 2752408, 2752410, 2708693, 2708204, 2743635, 2743637, 2763431, 2763434, 2730723, 2730722, 2780826, 2780819, 2767470, 2770625, 2782656, 2782657, 2651874, 2652963, 2746740, 2746749, 2746566, 2746569, 2749984, 2719896, 2764951, 2344131, 2745047, 2745049, 2749851, 2749602, 2750008, 2745021, 2777047, 2777034


Errors/Problems Fixed: 

		 2780959	"cifs share add" command cannot use ip option on 5.6RP1P2 environment.
		 2725972	Upgrade ftp rpm for including security fixes.
		 2776252	Changes for vxmon.
		 2771990	Command "nfs show fs" displays wrong error message.
		 2796103	Generate UMI code for 5.7P3.	
		 2777977	Upgrade vxvm to VRTSvxvm-5.1.132.215-5.1SP1RP2P2HF15_SLES11.x86_64.rpm.	
		 2805052 	Failed to switch normal mode to ctdb mode; getting CTDB core dumps.		 
2590189, 2779856	"vxfenadm -g" will be removed, change to use "vxfenadm -s".
2592820, 2740620    "NFS show fs" shows error "too many arguments".
2633743, 2633682    Provide a command to manually backup sfsdg configuration database.
2783403, 2783407    Make tcp tunables configurable. 	
2795588, 2795076	FileStore create fs show error information "Volume resources can"t be created".
2638340, 2637979	hwclock not syncing after node addition.
2767960, 2744096	Old domain information still present on GUI after changing the configuration of the FileStore cluster.
2752408, 2752410	Pick up Samba 3.5.14 for FileStore.
2708693, 2708204	Provide support for VCS IMF agents.
2743635, 2743637    Provide ability to import/export network configuration.
2809277, 2763413	SAV cannot detect the virus file from cab container file.
2763431, 2763434	SAV cannot handle mime container file, though the related decomposer engine is enabled.
2730723, 2730722	User cannot get error message even if command "fsppadm query -l" fails.
2780826, 2780819	Modify TCP parameters for performance.	
2761697, 2761702	snapshots have ctimes and mtimes listed in storage snapshot list that are days after the snapshot was created.
2431150, 2429200	"storage tier query" does not print the required argument "yes" on screen.
2780945, 2772235	While configuring export - Backup sfs script, it should also backup the authorized keys from the clients, so after upgrade, the nascli will work without having to set up the password free access again.
2767470, 2770625	Cluster reboot fails with file system not getting unmounted. File system should be force unmounted.
2782656, 2782657  	cli>cluster reboot all` with "ntp disabled", but ntpd is started automatically after reboot. It should be kept as "disabled".
2651874, 2652963	`antivirus set tmpfssize` needs improvements: 1) CLI does not show value, 2) PXEboot node does not have tmpfs mount, 3) Can change value while running realtime scan.
2746740, 2746749	ftp service is not restarted after pure-ftpd dies. Other services like NFS/CIFS are restarted.
2746566, 2746569 	msgs in sfsfs_event.log are not well organized when services stop/start and reboot. System should show which service has which status on each node. 
2749984, 2719896 	Executed nis show groups command under network mode. No memberships exist in the groups.
2764951, 2344131 	Backspace key should delete the previous password character.
2766437, 2766438    Two Web GUI summary pages should not truncate hyperlinks.
2745047, 2745049 	Cannot upgrade savjlu package correctly using rpm -U savjlu.rpm.
2749851, 2749602 	CLUSTER_ECHO: command not found.
2750008, 2745021 	crontab conf includes duplicated comments after modifying the snapshot schedule.
2777047, 2777034 	Disk format hangs if unable to get the controller details.


IV. KNOWN ISSUE:
----------------

Etrack incident: 2882635  

SYMPTOM: In FileStore 5.7 P3, if private interfaces are swapped using the System> swap command,
the LLT configuration does not get updated.

DESCRIPTION: When private interfaces are swapped using the System> swap command, using the lltconfig -a list command displays the old configuration.  This leads to private communication issues in the FileStore cluster.

RESOLUTION: Reboot the cluster.


Etrack incident: 2820537

SYMPTOM: When running the Storage> fs fsck command, you may encounter a process of Unknown.

DESCRIPTION: When using the CLI to run Storage> fs fsck, FileStore performs a normal fsck operation. For a normal fsck operation, there is no "pass0, pass1..., pass4" status; thus it is normal to display an Unknown status. Full fsck is run only if the normal fsck fails. In the support mode, if you are running a full fsck, FileStore records that status in an internal database/file.

RESOLUTION: The Unknown process is normal, since there is no process printed as output when running a normal fsck using Storage> fsck <fs name>.  If this is a full fsck, and it is recorded in an internal database/file, the process is displayed.

SYMPTOM: Upgrade hangs if the node has heavy I/O.

DESCRIPTION: Upgrade using SSH to install the VCS RPM on all the nodes. If the node has heavy I/O, the SSH connection will time out and the installation of the RPM will fail.  VCS will have a different version on all the nodes and the slave node will not be able to join the cluster. The upgrade will hang.

RESOLUTION: Manually upgrade the VCS RPM on all the nodes.
1. Extract SFSFS-patch-5.7P3_rc6_2012_06_13.tar.gz.
2. Run "hastop -local -force" on all the nodes. 
3. Run "rpm -Uvh cdcontents/symc/patch_pkgs/VRTSvcs-5.1.132.002-
SP1RP2HF2_SLES11.i686.rpm" on all the nodes.
4. Run "/opt/VRTS/bin/hastart" on all the nodes.

IV. NEW FEATURES:
-----------------

Etrack incident:  2633743 2794122

New support command: Support> vxconfig backup

In some cases, the sfsdg (FileStore disk group) may get corrupted, and it may need
to be reconstructed to a desired location. The Support> vxconfig backup
command allows you to manually backup the sfsdg configuration database to the
desired location. If you do not have the latest backup, then the configuration
database may not be restored to the latest changes.

Backup sfsdg information to the specific directory.

Syntax:
Support> vxconfig backup

Example:
Support> vxconfig backup /tmp
SFS vxconfig INFO V-288-0 Backup diskgroup(sfsdg) to /tmp complete.

Etrack:  2788094  2795850    

New replication commands for supporting replication of directory structures: 

Replication> set resync_options

This command sets the resync options of the job. The job resync command will trigger 
along with these options. Available options are  no_delete, only_sync_dir. By default, 
at the end of the job resync session, the job resync deletes files on the  destination
repunit  that  are not present on the source repunit. If you set the no_delete option, 
job resync does not delete  files  on  the destination  that  are not present on the source.
The job resync syncs files and directories of the source to the destination by default. 
If  you set the only_sync_dir option, job resync syncs only the source repunit directories 
to the  destination  cluster repunit; job resync does not sync files in the directories.  The
options are on a per replication job basis, and the  replication job  should  be  in  the  
Enabled or the Disabled state prior to changing these options.

The Replication> show resync_options 

This command displays the list of replication configurable parameters and their values.  These
are the settings that have been configured using the Replication> set command.  The Replication> show 
resync_options command displays the resync options of the replication job. 

Etrack: 2818048

Creation of a bond interface at FileStore installation time

An administrator can create a bond interface from a given list of public interfaces at FileStore installation time. This feature allows an administrator to save a number of physical IP addresses used
for installation and post-installation bond creation.

Please note this feature is about NIC bonding of public interfaces and bonding of private interfaces is not supported.

This feature includes an updated installation user interface with additional user inputs. An administrator can use the Help button to read the complete workings of the input installation screen. The installation screen has a shortcut for each input (Alt+Key). This reduces the use of the Tab key, and provides an easy way to navigate the installation screens.

The following are the two use-case scenarios for creating a bond interface at installation time:

1. The administrator does not want to create any bond interfaces and can continue with the normal FileStore installation.

- In this case, the administrator should not create any virtual bond device, and with valid inputs, the administrator can continue.
- The administrator does not require any changes in NIC list control of the yast menu.
- If you do not create any bond device (total device - 2 private - excluded device) will be considered for a public network.  The same applies to the virtual IP address (VIP) calculation.

2. The administrator wants to create a bond interface.

- To create a bond interface, click "Add New". The installation program prompts for the bond mode, and will select a unique bond name. The user can then add multiple virtual bond devices.
- To add a physical NIC to the bond interface, select NIC from the NIC List, (the user needs to check devices for selection from a checkbox menu of NIC List), and click "Add To".  This will provide a list of virtual bond devices that the user has added. Select one of the virtual bond devices, and click Ok. This will form an actual bond of given devices after installation. The user must select at least two devices for bonding. 
- After adding any virtual bond device, the NIC List will get updated with the new "bond#" members in it.
- The user can remove any added bond devices using the "Remove" button. The installation program deletes all references to physical NICs.
- For exclusion, select NICs from the NIC List control (note checkboxes) and click on "Exclude." The user does not need to find the PCD IDs and enter them manually.
- After adding physical NICs or excluding NICs from the installation program, the description in front of the physical device will change to "Slave of bond#", or in case of exclusion, it will display as "Excluded".
- This might be possible if the user wants to locate the actual hardware of the NIC. To do this, select physical NIC and click on "Identify". This will start LEDs blinking on the corresponding NIC with a definite duration ON/OFF pattern. 
- The user can re- "Include" the NIC that was previously "Excluded".
- Select Number of VIP per NIC using the Up/Down arrow keys.
- When the user inputs everything and presses "Next", then the installation process will do input validations and will report conflicts if any.
- After the user is done with the addition of the NIC for bonding and completes installation input (on Next), the installation process tries to find out private NICs from the list of eth* that are not part of any bond device and are not excluded. (Do not connect private NICs to the public network (that is, gateway IP)). No manual intervention is required for private NIC identification. 

IP address range calculation formulas:

- How to calculate the number of physical NICs:
Number of public NICs = NICs that are not part of any bond - 2 private NICs - 
number of NICs excluded
Number of physical NICs = (number of public NICs + number of bonded NICs) * 
number of nodes in the cluster.

- How to calculate the number of virtual NICs:
Number of virtual NICs = ((number of public NICs + number of bonded NICs) * 
number of nodes in the cluster) * number of VIPs per NIC.

The following inputs have the same meaning as those of the old inputs: 
Cluster Name, Physical IP Address Range Start, Virtual IP Address Range Start, Number of VIPs per NIC, Gateway, NTP server name, domain server name, Separate console port, Subnet mask