Date: 2013-09-18 OS: SLES OS Version: 10 SP3 Symantec FileStore 5.6 RP1 P4 HF6 Patch Upgrade README CONTENTS I. OVERVIEW II. UPGRADE PROCEDURE III. FIXES IN THE NEW PATCH IV. KNOWN ISSUES V. NEW FEATURES VI. APPENDIX PATCH ID : N/A PATCH NAME : SFSFS-patch-5.6RP1P4HF6_rc4_2013_07_30.tar.gz BASE PACKAGE NAME : Symantec FileStore BASE PACKAGE VERSION : 5.6 OBSELETE PATCHES : N/A SUPERCEDED PATCHES : N/A INCOMPATIBLE PATCHES : N/A SUPPORTED OS : SLES SUPPORTED OS VERSION : SLES 10 SP3 CREATION DATE : 2013-09-18 CATEGORY : enhancement, bug fix REBOOT REQUIRED : Yes SUPPORTS ROLLBACK : NO I. OVERVIEW: ------------ Symantec FileStore provides a scalable clustered storage solution. This document provides release information for the patch. II. UPGRADE PROCEDURE: ---------------------- After you have installed or synchronized a new Symantec FileStore patch into your cluster, the list of available commands may change. Please login again to the CLI to access the updated features. IMPORTANT: There is a downtime for the services during an upgrade. The actual downtime will be a little longer than it takes to reboot the system. To avoid data loss, Symantec FileStore recommends that customers stop I/O processing completely during a patch upgrade. After you apply this patch, you cannot uninstall it. The 5.6RP1P4HF6 patch can only be installed on 5.6, 5.6P1, 5.6P2, 5.6P3, 5.6RP1, 5.6RP1P1, 5.6RP1P2, 5.6RP1P3, 5.6RP1P4, 5.6RP1P4HF1, 5.6RP1P4HF2, 5.6RP1P4HF3, 5.6RP1P4HF4 or 5.6RP1P4HF5. If you are upgrading on a replication source cluster: 1) Pause running jobs. 2) Upgrade the cluster. 3) Resume all paused replication jobs. If you are upgrading on a replication target cluster: 1) On the target cluster, check "echo listfset | fsdb -t vxfs /dev/vx/rdsk/sfsdg/ | grep -w \"\" | wc -l" a. If the command returns 0. b. If the job is in the running state, pause the job from the source console. c. Go to step 3. 2) If the above command returns non-zero: a. If the job is in a running state, pause the job from the source console. b. Umount the checkpoint first using the command "cfsumount /vx/". c. Fsclustadm -v showprimary . d. Except for the file system primary node, run "hagrp offline -sys ". e. Once you have run "hagrp -offline" on the non-primary nodes, run "vxumount -f -o mntunlock=VCS " on the primary node. 3) Upgrade the cluster. After upgrading the online file systems that are offline in 2(b) and 2(d) steps, run the following commands after the upgrade: a. Run this command on all nodes, "hagrp -online -sys ". b. cfsmount /vx/. c. Resume all paused replication jobs. To install the patch: 1. Login as master: su - master 2. Start the patch install: upgrade patch install IMPORTANT: Full upgrade instructions are included in the Symantec FileStore 5.6 Release Notes. Please note the following revisions When you upgrade: Symantec FileStore recommends that you remove I/O fencing before upgrading any cluster node or exporting your current configuration. Use the Storage> fencing off command first followed by the Storage> fencing destroy command to remove I/O fencing. This step is not required, but it is suggested for a clean upgrade. III. FIXES IN THE NEW PATCH: ---------------------------- Etrack Incidents: 3021863, 3071059, 3132737, 3139904, 3142747, 3210003, 3210874, 3221252, 3226136, 3228847, 3247712, 3253887, 3254380, 3259734, 3259736, 3269900 Errors/Problems Fixed: 3021863 "smbd[29175]: disk_free: sys_popen() failed" message logged in /var/log/message many times. 3071059 `fsckptadm create` got frozen at vx_msg_send() on master. and `vx_msg_thread` got frozen at vx_recv_closeset()->vx_iget()->vx_ireuse() on slave. 3132737 `cli> storage snapshot quota on|off` may reset usage value incorrectly, this value may go to negative value easily like 18446744073709521067. 3139904 "storage fs shrinkto" does not finish even if it is performing for 24 hours or more. 3142747 nfs-lock(F_WRLCK) becomes no-effect when VIP is failed-over to other node. this should be kept. 3210003 `storage snapshot scheudle delete/create` mishandles FS-names with "-" hyphen, this seems to be one of hyphen-handling-issues. 3210874 file was saved as null file on quota file system. 3221252 `replication job enable job1` deletes line in crontab for "job10". this is one of name-mis-matching-issus. 3226136 if system has snapshot names "snap1" and "snap1_01", `cli>storage snapshot list` can't show right size for "snap1". 3228847 File names with special characters causes logging failure 3247712 data replication failed. 3253887 READ issue of the `reboot one of two controller on eternus` cause "fullfsck flag set". 3254380 `support debuginfo upload` for slave node, sfsfs_debuginfo*.tar.gz remains under working-dir on slave node, it does not remain for master node. 3259734 LLT peerinact value remains set as 18000 when debuginfo is performed simultaneously. 3259736 The LLT was not expired when LLT disconnected over 16 3269900 fullfsck flag was set in vx_attr_iget(). IV. KNOWN ISSUES: ----------------- Etrack: 3263117 Symptom: After `fs destroy FS`, fs_alert_conf still has fullspace-line for the FS, this should be removed at `fs destroy`. Description: fs alert full space information is not removed even after destroying fs. Resolution: Please ensure that alert is turned off before destroying fs. Otherwise fs alert information won't be cleared. Etrack: 3244325 Symptom: Scheduled relocation job failed "No such file or directory" for the destination server. Description: This issue may occur if a repunit is defined as a sub-directory in a file system and there are directory operations (add, delete, move, and so on) happening inside the repunit This issue does not occur if the user defines the repunit using the entire file system. Resolution: The workaround is to define the entire file system as the repunit. Etrack: 3285662 Symptom: replication process "rep_rm_src_star" dumps core when FileStore has an excessive amount of metadata changes, for example, over 10 million. Description: This is a by design "out of memory" issue. Incremental replication cannot handle more than approximately 20 million non-rename FCL changes or 10 million rename changes due to the 4G virtual memory limit by the 32-bit library. Resolution: The workaround is to set the replication schedule to trigger the job more frequently to avoid too many changes at one time for an incremental replication session. If the failure happens, perform a "resync" on the job, and trigger the job again with shorter intervals. V. NEW FEATURES: ------------------ N/A VI. APPENDIX: ------------------ N/A