Should be sticky: Samsung 840 and 840 pro are not LSI megaraid compatible

Status
Not open for further replies.

0egp8

New Member
Apr 9, 2013
11
0
0
There is actually another firmware that was released on 5/30/13 (23.16.0-0012). I've tried both with 23.16.0-0012 & 23.12.0-0013 (which was released on 5/7/13). I can't get the diskcache to be enabled. It's just grayed out in the WebBIOS. I even tried with Cachecade trial key hoping that it'll make a difference but no go...

Are they enabled by default or did you have to choose "enable" from like a drop down menu?
You're right, my head skipped over the 16 in your firmware revision number. Can you check if the disk cache is currently enabled? You can do so by running

Code:
MegaCli -LDGetProp -DskCache -Lall -aALL
 

abackbone

New Member
May 30, 2013
2
0
0
Australia
Update your 9271-4i to the latest firmware (23.12.0-0013), and you should be fine.



Make sure that there's adequate airflow across the LSI card's heatsink (at least 200 LFPM as per spec). Temps can be outrageously high without them; I've gotten as high as 94C idle with bad placement, and the climate here is cool. There may be temperature throttling under load, which would botch the results. I'll test whether cooling makes a difference on my 9271-8iCC after some 25mm fans arrive.
Yep, I have seen this also, so we always return the server to the nicely air con'd room before we stress test.
Ambient 22, case 26 and the max I have seen our controllers go to is 60ish Celsius. We use Supermicro / Maxtron server stuff and the axial flow arrangement is quite good. we even run cooling on balanced.

One day I will tell you of the awesome rig I designed and was subsequently shafted by all and sundry (LSI). Quick specs, Multi card load balancing raid on dual chipset mainboard. Forced figures with 30 x Seagate 2TB 7200rpm SAS and 6 x Intel X25-E 32GB were in the neighbourhood of 2.35GB/s sustained read and write. But the rig resets and disables the clustering features on every reboot. Bummed.

And now PMS card. Singular. So sad.

Testing begins tonight.

Oh, something else I discovered with the new View software. I am told there are issues with non local admin accounts. IF you are logged in as someone else, get the software installed, let the password section fail, just copy the address from the shortcut, "Browser" - ' run as another user' and run it as the local admin, then browse to the address of the management sw.

Another hour in the bin for that one.

Back later.
 

mrkrad

Well-Known Member
Oct 13, 2012
1,244
52
48
the 9266 should work fine with the newest dx05 - well as fine as you want to call it. Basically all SSD fail under heavy load. Same with the 7 series maxcache.

fact is you need to overprovision your ssd - alot. 28-30% is real man. All the problems go away.

(the dx05 firmware seriously just turns the light on, if you think there is anything more you are smoking da good stuff, I asked.)
 

BigBuild

New Member
Jun 7, 2013
2
0
0
9285-8e and 9266-4i

I have a pretty big rig with multiple LSI Raid cards. Among them are an 9285-8e and a 9266-4i. The 9266 has 4 * 840 Pro 256 GB and the 9285 has 8 * 840 120 GB. I have the most recent firmware on all the SSD's and all controller cards and still can not enable Disk Cache in MSM. Mobilenvidia posted to this thread with an option that changes the enabled functionality in MSM using MegaSCU, but I can not find the utility MegaSCU anywhere. Can someone please tell me where I can download MegaSCU. I also have a theory. I think it is possible, perhaps even probably LSI has done this intentionally to magnify the the performance advantages of their Fast Path and Cache Cade software. If that is the case I don't have a problem paying for the theses software options, but I'd like to know if anyone has used Fastpath with Samsung 840 SSD's?
 

mrkrad

Well-Known Member
Oct 13, 2012
1,244
52
48
Sure fastpath is great, just not the 840 pro. Works great with the 830
 

BigBuild

New Member
Jun 7, 2013
2
0
0
Do you know where to download the utility MegaSCU? Also, have you been able to successfully make the mods posted by Mobilenvidia
 

bvbmedia

New Member
Jun 11, 2013
4
0
0
I've bought a LSI raid card purely because LSI advertises that their cards are perfect for SSD. But with this card I can't enable disk cache. The option is grayed out and showing "unchanged".

Controller: LSI MegaRAID SAS 9271-8i with firmware: 23.12.0-0013, 3.240.25-2382
SSD's: 8x Samsung 840 Pro with firmware: DXM05B0Q

Please LSI fix this annoying issue.
 

bvbmedia

New Member
Jun 11, 2013
4
0
0
Looks like LSI is planning to publish firmware MR 5.7 for MegaRAID SAS 9271-8i:

MegaRAID SAS 9271-8i

The MegaRAID MR 5.7 firmware will be posted shortly, please check back on this site for updates
 

mrkrad

Well-Known Member
Oct 13, 2012
1,244
52
48
The issue is tackled using the following methods:

megascu to enable caching.
add battery or cachevault.
enable write back caching (!!)
overprovision to 30% ie 192gb from 256gb, avoid any models less than 256gb. avoid non-pro.
change patrol scan and consistency check to 1% instead of 30%
Ensure battery is not mounted on card or it will overheat. (or just skip the battery).
dx05 firmware.

The biggest changes have been from doing a secure erase, OP to 30%, PATROL/CC to 1%, FULL INIT of all sectors in foreground (Takes minutes).

Knock on wood but so far so good.

I also moved from raid-10 to raid-1 with extent spanning, with vmfs-5, it appears to scale in random iops better than a big raid-10. STRIP SIZE is very important, if you are running a 64KB strip, then you are killing latency since a write may cause a P/E against 64KB per DRIVE (STRIP PER DRIVE, STRIPE PER ARRAY)!!.

Think about 1 byte changes that cause a 64KB strip P/E versus 8K (16K For modern M500 1TB 16KB PAGE SSD).

Linear speed tanks, but random latency flattens out.

Otherwise the P420 $250 1gb/fbwc with 100% write (1gb) and 0% read ahead dampens out the issues (OP still).

Right now i'd have to recommend buying the intel S3500 datacenter drives, given the $900 for almost 900gb and it requires only 15% OP and has capacitor, I would say ditch LSI junk (except software raid) and stick to the new affordable hotness from Intel.

Whilst I've had more failures from intel drives than any other, it's probably because we've also got 80+ operational intel drives versus 30 samsung.

The free next day air advance swap (intel partner inside!) is far better than the samsung rma process which gives no freebies to partners even though we buy and sell the piss out of their drives. lol.

no 4K sector support? come on seriously. Hello welcome to 5 years ago when the first SSD came out!
 

bvbmedia

New Member
Jun 11, 2013
4
0
0
Thanks for sharing the direct link with us!

Here is the list of changes in MR5.7:

Bug Fixes and Enhancements:
===========================
FIRMWARE:
**NOTE: Full native 4K sector Firmware support is deferred to MegaRAID Release 5.8 targeted for September 2013.
SCGCQ00414056 - (Closed) - Potential IO integrity with R0/R1 CIO WT when overlapping R/W issued to LD.
SCGCQ00413472 - (Closed) - Unable to perform Consistency Check and other background operations in a legacy WebBIOS
SCGCQ00413254 - (Closed) - Multiple VDs undergoing full init show a calculated estimated time remaining even though they sit at 0%
SCGCQ00406688 - (Closed) - 'Static Code Analysis Tool' warnings: IDs - 10007 , 10338 , 10340, 10356, 10364 , 10371, 10389, 10390, 10847, 10848
SCGCQ00406301 - (Closed) - 'Static Code Analysis Tool' warnings 10071, 10100, 10162, 10174, 10288, 10332, 10361, 10363, 10503
SCGCQ00406219 - (Closed) - FW reports "Progress init not proper" during 16 VD Rebuild
SCGCQ00400564 - (Closed) - Delete VD while PR still in progress cause debugger to loop. Reboot the system, debugger still loops
SCGCQ00393749 - (Closed) - MonTask: line 281 in file ../../raid/mem.c if running megacli AdpDiag
SCGCQ00393117 - (Closed) - Firmware crash during CME recovery
SCGCQ00392421 - (Closed) - 5.7‘Static Code Analysis Tool’ changes
SCGCQ00385372 - (Closed) - 'Static Code Analysis Tool' Defect OVERRUN STATIC: CIDs Out-of-bounds access
SCGCQ00369611 - (Closed) - Code Analysis Tool fix
SCGCQ00349837 - (Closed) - Sequential Read performance is less on CacheCade associated LD's.
SCGCQ00424928 - (Closed) - Application could not make a JBOD bootable without a VD present
SCGCQ00412766 - (Closed) - In HII, when user has a critical message and the conrtoller is in Safe Mode, going into CTRL Management will result an error msg
SCGCQ00404301 - (Closed) - Significant drop in Sequenial read performance with Read Ahead policy
SCGCQ00404263 - (Closed) - Staggered CMEs across R10 array are causing a read error
SCGCQ00384874 - (Closed) - MR5.6 Beta Firmware BST: Code Analysis Tool Fix
SCGCQ00378971 - (Closed) - Continues PR 2nd time on Raid 1 and then delete VD and reboot system cause debugger to continues loop and want to run PR
SCGCQ00378358 - (Closed) - 5.7 FW PKG 23.11.0-0013/Using CLI, can change a cachecade PD to WB even though enableSSCWB=0
SCGCQ00422757 - (Closed) - Firmware does not progress background operations with certain systems when using WebBIOS in legacy mode
SCGCQ00406148 - (Closed) - Grammatical spacing error in bios message at POST
SCGCQ00400555 - (Closed) - The elapsed time shown in the FW debugger for a reconstruction not starting at 0
SCGCQ00398281 - (Closed) - "Question value mismatch with Option Value" message in HII's Controller Management when in Safe Mode with pin cache
SCGCQ00394067 - (Closed) - Removed unwanted prints. For copyBack, total Blocks for copyback was not getting intitailized for right drive. Initialized for copyBack destination PD. Added necessary fix.
SCGCQ00383251 - (Closed) - Code Analysis Tool Defect OVERRUN STATIC: CIDs Out-of-bounds access
SCGCQ00381562 - (Closed) - Assertion failure fixed for issuing "adpdiag" from megacli-dos
SCGCQ00378408 - (Closed) - Correct formatting and compilation issues
SCGCQ00400149 - (Closed) - Firmware failed to import drives to start rebuild after remove then reinsert
SCGCQ00394259 - (Closed) - Fixes for defects found by Static Code Analysis Tool
SCGCQ00393347 - (Closed) - 'Static Code Analysis Tool' code changes
SCGCQ00383337 - (Closed) - Firmware is sending totalElapsedSecs value as 0 for LD and PD progresses
SCGCQ00380450 - (Closed) - MegaCli pdFwDownload fails on a SATA drive with an interposer (parent defect is SCGCQ00352210)
SCGCQ00378916 - (Closed) - Timed critical boot message under UEFI mode not showing
SCGCQ00375763 - (Closed) - Fix for Firmware asserted when VDs are offline
SCGCQ00375418 - (Closed) - Code Analysis Tool Fix
SCGCQ00393765 - (Closed) - Code Analysis Tool Fix
SCGCQ00391493 - (Closed) - Firmware blocked firmware download to interposer
SCGCQ00372058 - (Closed) - NCQ support is disabled in FW
SCGCQ00334864 - (Closed) - Double media error getting corrected during check consistency
SCGCQ00406233 - (Closed) - Data Miscompare found while running I/O overnight with degraded virtual drives
EnhancementRequests (12)
**NOTE: Full native 4K sector Firmware support is deferred to MegaRAID Release 5.8 targeted for September 2013.
SCGCQ00323433 - (Closed) - 4K sector support - Cache Mgmt, Snapshot, LD config, RAID **
SCGCQ00369610 - (Closed) - CME handling improvements for handling multi-row I/Os
SCGCQ00369835 - (Closed) - 4K sector support - Switch to enable/disable 4K feature, Restrict "Snapshot" to 512-sector drives **
SCGCQ00392264 - (Closed) - Firmware returns invalid data in OOB packet
SCGCQ00369206 - (Closed) - ECC memory error boot message and event log enhancement.
SCGCQ00369208 - (Closed) - Rebuild time reset
SCGCQ00392280 - (Closed) - Provide Fastpath as a standard offering for TB based controllers
SCGCQ00364773 - (Closed) - Ability to power off and on individual HDD
SCGCQ00369741 - (Closed) - Port latency outlier fixes to FW (from legacy PR137246)
SCGCQ00323427 - (Closed) - 4K sector support - DDF, DM, Host Interface **
SCGCQ00369832 - (Closed) - 4K sector support - iMR, analyze all remaining hard-coded instances of "512" **
SCGCQ00369228 - (Rejected) - RRB 45 - CLI Reporting of TMM VPD
CSETActivities (52)
SCGCQ00354181 - (Port_Complete) - DDF rev is not updated properly when both 4K and 512 drives are present **
SCGCQ00372172 - (Port_Complete) - MR2208 JBOD function can't handle the UNC error of SATA disk
SCGCQ00381798 - (Port_Complete) - MR Controller Rejects Vendor Unique SCSI Command
SCGCQ00395186 - (Port_Complete) - Double media error getting corrected during check consistency
SCGCQ00396296 - (Port_Complete) - Ctrl+R- Ctrl+R BIOS hangs while hot plug of 2 enclosures with more than 32 drives
SCGCQ00397246 - (Port_Complete) - [MR5.4] CacheCade: ECC/Medium/Unrecoverable read errors handling Enhancement
SCGCQ00397338 - (Port_Complete) - CC2.1_RA_Support : Data Integrity issue may happen upon creating WB CCVD during Cache Flush
SCGCQ00397347 - (Port_Complete) - Sequential Read performance is less on CacheCade associated LD's.
SCGCQ00414020 - (Port_Complete) - Adjusted the USB phy boost value for the OEM controllers with SubDevice IDs between 3510 and 3515.
SCGCQ00414463 - (Port_Complete) - IO integrity error seen after Source VD associated with CacheCade VD is forcibly unblocked
SCGCQ00414466 - (Port_Complete) - During rebuild on PRL11 CacheCade VD and rebuild on RAID6 Source VD IO integrity error is seen.
SCGCQ00417742 - (Port_Complete) - BBU Charging gets disabled when Transparent Relearn fails to start due to low capacity
SCGCQ00373301 - (Port_Complete) - SCSI Pass-Through to SAS SSD returns data but disk reported medium error
SCGCQ00374850 - (Port_Complete) - Foreign config import failed from MR to iMR on Solaris 11x86
SCGCQ00384609 - (Port_Complete) - Assertion failure in ../../raid/raidpci.c at line 10011 , while running IO with medium errors on arrays.
SCGCQ00385362 - (Port_Complete) - reading SATA IDENT information using STP_Passthru doesn't work with latest FW
SCGCQ00389234 - (Port_Complete) - Continues PR 2nd time on Raid 1 and then delete VD and reboot system cause debugger to continues loop and want to run PR
SCGCQ00395380 - (Port_Complete) - Disk missing sporadically during reboot of OEM
SCGCQ00405889 - (Port_Complete) - Auto Foreign import fails when controller is booted with pinned cache.
SCGCQ00406635 - (Port_Complete) - Power loop cycle test on a non DIF drive connected to LSI controller caused fatal firmware error
SCGCQ00411291 - (Port_Complete) - Significant drop in Sequenial read performance with Read Ahead policy
SCGCQ00414469 - (Port_Complete) - When a Cachecade PD marked as missing is reused to make Cachecade VD IO integrity error is observed.
SCGCQ00418110 - (Active) - If the hotspare was spun down, the copyback after a drive failure wouldn't always start.
SCGCQ00381325 - (Port_Complete) - Wrong event generated "Reminder: Potential non-optimal configuration due to drive PD xx(e0xfc/sx) commissioned..
SCGCQ00383643 - (Port_Complete) - "Assertion failure in ../../raid/mem.c at line 281" on issuing "adpdiag" from megacli-dos
SCGCQ00392859 - (Port_Complete) - TFM modules with dirty data can cause the controller to hang during boot when transported to another controller
SCGCQ00409501 - (Port_Complete) - The reliability and error recovery of the SuperCap firmware was improved.
SCGCQ00418100 - (Port_Complete) - If a correctable ECC error occurred during a cache offload, the CPU would crash, resulting in a data loss
SCGCQ00360907 - (Port_Complete) - VD becomes WT with a good & new Super Cap on boot
SCGCQ00381814 - (Port_Complete) - (multiple) Megacli64 error in deadlock state
SCGCQ00386772 - (Port_Complete) - Timed critical boot message under UEFI mode not showing
SCGCQ00392416 - (Port_Not_Required) - Firmware blocked firmware download to interposer
SCGCQ00393320 - (Port_Complete) - NCQ support is disabled in FW
SCGCQ00397337 - (Port_Complete) - Data Integrity issue may happen upon creating WB CCVD during Cache Flush
SCGCQ00412657 - (Port_Complete) - Adjusted the USB phy boost value for the OEM controllers with SubDevice IDs between 3510 and 3515.
SCGCQ00374855 - (Port_Complete) - GHS drives changed to unconfigured good after OCR / System reboot scenario's
SCGCQ00378050 - (Port_Complete) - BMC see clock stetching on I2C bus
SCGCQ00380935 - (Port_Complete) - BMC see clock stetching on I2C bus
SCGCQ00381794 - (Port_Complete) - MR Controller Rejects Vendor Unique SCSI Command
SCGCQ00391703 - (Port_Complete) - Copyback was successful when the block sizes did not match
SCGCQ00397250 - (Port_Complete) - [MR5.4] CacheCade: ECC/Medium/Unrecoverable read errors handling Enhancement
SCGCQ00405886 - (Port_Complete) - FW halts when you have pinned cache and remove VD form the configuration.
SCGCQ00374852 - (Port_Complete) - Defect fix for 4 K drives Foreign configuration import.
SCGCQ00409495 - (Port_Complete) - The reliability and error recovery of the SuperCap firmware was improved.
SCGCQ00412679 - (Port_Complete) - Timeout when creating a hotspare
SCGCQ00385361 - (Port_Complete) - reading SATA IDENT information using STP_Passthru doesn't work with latest FW
SCGCQ00395381 - (Port_Complete) - Disk missing sporadically during reboot of OEM
SCGCQ00354097 - (Port_Complete) - MR: VD created in IR controller comes as UG when connected.
SCGCQ00380285 - (Port_Complete) - Defect fix for 4K configuration record updation is not happening with 4k drives **
SCGCQ00381728 - (Port_Complete) - Firmware asserted at raid/cache.c at line 3661 when VDs are offline
SCGCQ00381812 - (Port_Complete) - (multiple) Megacli64 error in deadlock state
**NOTE: Full native 4K sector Firmware support is deferred to MegaRAID Release 5.8 targeted for September 2013.

WebBIOS:
SCGCQ00380528 - (Closed) - In EFI WebBIOS with CC running on a R1 when 1 PD is removed EFI WebBIOS still shows that CC is in progress
SCGCQ00400124 - (Closed) - Controller breaks into Megamon when clicking on the enclosure link in WebBIOS

NVDATA:
SCGCQ00333522 - (Closed) - Degraded performance comparison With the latest MR 5.3 -MR_3.190.25.1887

Hii:
SCGCQ00420373 - (Closed) - HII doesn not allow user to use the remain free capacity to configure additional VD for R50/R10.
SCGCQ00422774 - (Port_Complete) - MR 6.1 HII: empty warning prompt is given when creating a CacheCade VD with forced writeback option
 

bvbmedia

New Member
Jun 11, 2013
4
0
0
When enabling the disk cache without further tuning looks like asking for troubles:

0egp8 wrote on http://userpart.com/webhost/showthread.php?t=1223799&page=6

Did some testing with the disk cache enabled. Enabling the disk cache makes SSDs drop from the array under load. Happened in two separate instances with two different Samsung 840 Pros with nearly the same sequence of events.
Seems all we can do is be vocal about it to LSI.


mrkrad, till now you did not have have any serious issues?
 

yabo

New Member
Jul 23, 2013
1
0
0
The link to the 5.7 download does not appear to work anymore.

I downloaded 5.6, but am I correct to assume the problem is not fixed in that version? My rates with x8 128GB Samsung 840 Pro's still seem very very slow.
 

mrkrad

Well-Known Member
Oct 13, 2012
1,244
52
48
Works fine I run many servers with this. MEGASCU disable the cache blocking feature, set the patrol scan and other scan from 30% to 1%, secure erase (use a desktop) the drives, format to 30% OP (192gb out of 256gb). Solid. Zero problems.

I have 6 servers with LSI 9260 series and 840 pro and 1 DL380 with dual 9266 to 840 pro 512gb.

do not use the 128gb models. the 512gb @ 384 gb is solid.

I guarantee if you do not use 30% OP you will have drops.

Stupid fast! Had I to go back, I would have done 4 controllers (x4) on the dl380 g7 rather than 2 controllers !

ESXi is faster when each vm has its own controller, its own drives.

Lack of SR-IOV support is the problem!
 

mrkrad

Well-Known Member
Oct 13, 2012
1,244
52
48
This is no longer the case, the 840 pro is now supported by LSI for all cards it appears. I retract the statement!
 
Status
Not open for further replies.