Professional Documents
Culture Documents
Tunesys-Datarebalancing Checklist v2
Tunesys-Datarebalancing Checklist v2
Pre
1 Access Permission
2 Remote Access
3 Opening Email
4.a CPG
5 Logs Capture
6 System Check
7.a HW Rebalance
8 OS
9 OS
10 DO License
11 Dedup
12 Dedup
14 AO
15 Active Tasks
16 Drive Readiness
18.a PD Health
18.f PD Utilization
Exe
Note :- Only one instance
2 Tunesys execution
3 Task Validation
Pos
1 Task Completion status
2.a
3 Dedup
4 Resume Tasks
5 Uninstall Scripts
Addition
1 Alternate Commands
Cancel T
1 Cancel Tunesys
Tunesys / Data Rebalancing
Pre Checks
Validate if we have permission from the customer to execute tunesys. If not please ask for permission. CHECK SSI.
Check if remote access is available via RDA/NGDC and confirm with the customer if this can be performed remotely or v
Create a Task for tunesys in PSA and update it on day to day basis. This can be closed post completion of final data rebal
Check if CPG Parameter modification is a PRE-REQ in CSR Technical Write up. Perform analysis from Sales upgrade, show
the activity.
Validate if previously on any trip / phase if CPG parameters were changed to meet the requirement for a data movemen
Validate the Storeserv s/n, to make sure we are working on correct system. CLI # showsys -d
Validate all the trips / phases are completed by checking the HW configuration from Config file or from the system again
Also please check if there was a revised config approval if any.
Ensure that HW is evenly balanced across the node pairt such as having same no of drives / cages behind each node pair
data rebalancing phase.
If the HPE 3PAR OS is 3.2.2 and above, tunesys starts automatically after HW Upgrade (post admithw). Please verify if th
CLI#showtask -active or CLI#showtask.
For any version of 3.1.1 OS Family, tunesys requires DO License. Please verify using command: CLI# showlicense
From HPE OS 3.1.2 onwards, tunesys does not require DO License.
Note: DO NOT RUN TUNEVV ON DEDUP VOLUMES IF THERE ARE GREATER THAN 16 TDVV IN A CPG ON SYSTEM BELOW
OUTAGE.
On DEDUP ENABLED ARRAY, install below scripts as per below simplified table pruor to tunesys execution
Ex to identify dedup volumes from CPG:- ----- Inserv CLI # showcpg -----
--------------------(MB)---------------------
----Volumes---- -Usage- ------ Usr ------ ---- Snp ---- ---- Adm ----
Id Name Domain Warn% VVs TPVVs TDVVs Usr Snp Total Used Total Used Total Used
Validate if the system has dedup and compression enabled by running the command # showvv -s
Validate version of TDVV by running the command # showcpg -d
For TDVV3 (3.3.1.x onwards), #showcpg -d <CPG name> will display the TDVV version, sample output below. Please ensu
TDVV2, you may not have the version # displayed
Check the free space available in the array using command showspace. Tunesys task requires certain amount of free spa
total free space in each tier. The probability of tunesys failing is high when the array is over 85% utilized. The array over
AO should be suspended before Data Movement Activity, take permission from customer before execution.
Please check if the Adaptive Optimization is licensed and configured by Issuing CLI # showaocfg and
If AO Policies are defined (showaocfg Output), suspend all scheduled tasks using command
CLI# setsched -suspend_all
Note:- Any scheduled tasks that involves region moves has to be suspended [mandatory].
Issue command CLI # showtask -active and wait for any running task to complete including AO Task.
It is OK to have remote copy sync running as an active task.
Please execute tunesys only after ensuring chunklet initialization is complete on newly added drives. Issue command
CLI# showpd -c to verify (Un-initialize chunklet count should be 0)
If the activity includes addition of node pair + cages + PD's behind the new node pair only, Please check the free space o
or above, we need to call out to the customer that once node pair is added, a temporary CPG should be created behind
space behind existing node pair prior to tunesys execution.
Command to check the free space behind node pair Tier wise: cli# showpd -c -p -devtype <FC/NL/SSD> -nd 0,1
Caution!!! Failing to check this may lead to an array outage
Template: "Dear Customer - Since the current (Array or Tier) utilization is 90% and above post HW upgrade please create
some VV's to free up space behind existing node pair.
Reason: If one of the node pair runs out of space, new LD creation may not be possible and CPG may go to a failed state
Ensure that all drives are in normal state. Data evacuation should not be performed when the drives are in degraded / f
verify.
CLI# showpd -failed -degraded ==> None to be listed as failed or degraded
CLI# showpd -s ==> There are two status, The main state and detailed state and both should be normal.
Sample Output below with normal status
----- showpd -s -----
Id CagePos Type -State- -Detailed_State-
0 0:0:0 FC normal normal
1 0:0:1 FC normal normal
Perform Healthcheck on the array and ensure that array is healthy with no show stoppers prior to execution of tunesys
(Send details to swd-3par-gss-ssc@hpe.com )
If healthcheck reports LD's are degraded Ex:- From ha cage to ha mag or LD's chunklets on a remote disk, this may get re
available free space.
Please reach out to Escal team for further clarification.
23 3PAR-Vmware-Test-LUN23 3par-esx10 0:1:1 t 1466 1466 1466 93544 93544 93544 5.75 5.75 63.8
24 3PAR-Vmware-Test-LUN24 3par-esx10 1:1:2 t 1435 1435 1435 91461 91461 91461 5.46 5.46 63.7
If the service time is over 25ms, this is an indication of host latency and tunesys should not be initiated.
Check the performance of the PD prior to the execution of DO. Verify by issuing the below command
CLI# statpd -devinfo -iter 1
Sample Output below
13:14:52 05/11/2016 r/w I/O per second KBytes per sec Svt ms IOSz KB Idle %
ID Port Cur Avg Max Cur Avg Max Cur Avg Cur Avg Qlen Cur Avg
0 1:0:1 t 84 84 84 7366 7366 7366 7.73 7.73 87.9 87.9 1 60 60 ==> FC
1 0:0:1 t 47 47 47 1690 1690 1690 5.22 5.22 3 6.2 36.2 0 85 85 ==> FC
- NL PDs maximum IOPs is 75 @ 8K reads
- FC 10K PDs maximum IOPs is 150 @ 8K reads
- FC 15K PDs maximum IOPs is 200 @ 8K reads
If the IOPS is more than the above mentioned threshold, please check with Escal and notify to the customer that the arr
read/write performance.
NOTE:- Please ensure to capture session logs
Using historical data on STaTS, verify data from 7 days perform.db file whether delack counter has increased. This would
assist customer reach support to address the performance issue.
Page Statistics
------------CfcDirty------------ --------------CfcMax--------------- ---------
Node FC NL SSD_150KRPM SSD_100KRPM FC NL SSD_150KRPM SSD_100KRPM FC N
0 962 8 0 4969 110400 7200 0 81600 0 162960
1 2126 26 0 5252 110400 7200 0 81600 0 175501
2 1153 104 0 6386 110400 7200 0 81600 0 208356
3 897 14 0 6182 110400 7200 0 81600 0 192514
Running compactcpg with >=8TB fully provisioned VV can result in sysmgr hang.
This issue can be experienced when a compactcpg is initiated for a CPG which holds a fully provisioned VV sized >= 8TB
Due to an issue in code (variable overflow) sysmgr will start spinning in an infinite loop with the lock still outstanding. As a resu
makes the array un-manageable.
Fix: Patch P92 (322 EMU4) | 322 MU6 | 331 MU1 onwards
Under provision type it will be listed as Full, if the FPVV size is more than 8TB, please hold free space reclamation until P92 for
Check if the system has Flash cache and if SSD's were added recently.
If yes, please inform customer to take a maintenance window to delete the existing Flash cache configuration and recre
tunesys cannot tune flash cacheLD's. If customer is not aware of this procedure, engage breakfix team by calling 1800 63
customer & support team and get them in conference. And this has to be documented in an e-mail which includes the cu
Ex to identify flash cache LD's using CLI # showflashcache
Issue command CLI#showcpg to identify the number of CPG's and their virtual volumes. CLI# Showcpg -sdg com
CPG configuration.
Execution
Note :- Only one instance of tunesys can be run at a time.
Perform a dry run on the array to assess the no of tunes suggested.
Issue Command: CLI # tunesys -dr
Note:- If dry run does not suggest any tunes, inform customer that we will do the analysis and execute suitable tune com
*** Keep customer informed ***
Note:- If dry run does not suggest any tunes, inform customer that we will do the analysis and execute suitable tune com
*** Keep customer informed ***
Tunesys runs as an active task and can be verified using below commands.
CLI # showtask -active
CLI # showtask -d <tunesys task ID#>
Sample Output below:
cli% showtask -active
Id Type Name Status Phase Step -------StartTime-------- -FinishTime- -Priority- -Use
26006 system_tuning tunesys active 2/3 102400/15126528 2016-05-10 10:36:59 AEST - n/a 3par
26007 background_command tunenodech active 2/2 100/7402 2016-05-10 10:42:23 AEST - n/a 3par
Share the tunesys task ID and inform that we are monitoring this task remotely and add the below note to the e-mail th
Please share the overall PD utilization of the array to compare pre and post tunesys changes with an existing template
Note: If the system is licensed and configured for Remote Copy and when the tunesys task is active and if a fail over occu
CLI# canceltask <TASK ID> (for more details refer cancel tunesys task sectio
Post Checks
Ensure that tunesys is completed by verifying the task details especially, the no of tunes performed and no of tunes succ
CLI # showtask -active (Tunesys task should not be listed if its already completed)
CLI # showtask (Make a note of the tunesys task ID and ensure its in done status)
CLI # showtask -d
Copy showpd -c output to an Excel sheet and covert this data from text to columns. Add a new Column (PD Utilization) a
Chunklets * 100
Filter the Drive type along with different Capacity and check the PD Utilization percentage. It should not be more than a
course of action.
al OK Fail Free Uninit Unavail Fail OK Fail Free Uninit Fail PD Utiliz
9 726 0 41 0 0 0 0 0 52 0 0 88.6446
If the HPE 3PAR OS is 3.2.1 and above version that has dedup enabled, we have to run a Dedup Garbage Collection scrip
Dedup Script and Procedure
Ex to identify dedup volumes from CPG:- ----- Inserv CLI # showcpg -----
--------------------(MB)---------------------
----Volumes---- -Usage- ------ Usr ------ ---- Snp ---- ---- Adm ----
Id Name Domain Warn% VVs TPVVs TDVVs Usr Snp Total Used Total Used Total Used
Post completion of the tunesys activity, resume all scheduled tasks. CLI#showsched -resume_all
Uninstall (relevant scripts) and remove all scripts post execution of the activity
Inform the customer / HPE whoever is the point of contact that this CSR is completed and we will go ahead and close the
Team PDL 3par.sps.csr@hpe.com
Additional Commands
If tunesys does not balance the data even after task completion, the alternate option is to run the below command prov
If the dry run is suggesting more tunes, safely proceed and execute the below alternate command. Consult CSR Team / E
CLI # tunesys –nodepct 1 – chunkpct 1 (All OS versions)
CLI # tunesys -tunenodech –chunkpct 1 (From 3.2.1 and above OS)
CLI # tunenodech -node <node ID> (From 3.2.1 and above OS)
CLI # tunesys -tunenodech -devtype <FC/SSD/NL> (From 3.2.2 and above)
CLI # tunenodech -devtype FC -node 0 (From 3.2.1 and above).
Tunenodech has good success rate if run individually on nodes than if ran within th
NOTE:- Post task cancellation verify status using the command CLI# showtask -active and if there are any activ
subtasks initiated by tunesys (maintask).
Sample Below.
root@1111111-1 Fri May 1 10:27:22:/common/support# showtask -active
Id Type Name Status Phase Step -------StartTime-------- -FinishT
26006 system_tuning tunesys active 2/3 2149376/15126528 2016-05-10 10:36:59 AEST -
26209 move_regions tuneld: tp-1-sd-0.86 active 1/1 933/1024 2016-05-12 10:11:42 AEST -
root@1111111-1Thu May 1 10:42:02:/common/support# canceltask 26006
Are you sure you want to cancel task 26006?
select q=quit y=yes n=no: y
root@1111111-1 Thu May 1 10:42:27:/common/support# showtask -active
Id Type Name Status Phase Step -------StartTime-------- -FinishTime- -Pr
26208 background_command tuneld active 1/1 0/1 2016-05-12 10:11:39 AEST - n/a
26209 move_regions tuneld: tp-1-sd-0.86 active 1/1 966/1024 2016-05-12 10:11:42 AEST - n/a
root@1111111-1 Thu May 12 10:44:46:/common/support# showtask -d 26006
Id Type Name Status Phase Step -------StartTime-------- -------FinishTime------- -Priority-
26006 system_tuning tunesys cancelled --- --- 2016-05-10 10:36:59 AEST 2016-05-12 10:43:29 AEST n/a
Note:- Initiate Insplore for analysis *** Keep Customer Informed ***
cklist
Links
CSR Sharepoint
CSR Sharepoint
Run the script disable_dedup_GC to disable/enable the dedup GC using
U21 kit.
Advisory
https://www.storefrontremote.com
Dedup Script & Procedure Download
Procedure to capture logs
Task status
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Comments for Pending and NA