kahlif
Contributor
Contributor

ESX 4.1 and LSI Megaraid Storage Manager

Jump to solution

Hello everyone!

I am testing a move from VMWare Server to vCenter for my organization and I am running into a problem.

In my test environment, I have a Supermicro H8DME-2 motherboard with an LSI 9261-8i SAS Raid controller on it. I currently have 2x750GB Sata drives in a Raid 1 configured on the LSI controller.

Now, the health in vSphere reports everything perfect, including degraded when I pull one of the hard drives.

My problem is my ability to rebuild the array if I replace a drive. I do not see that functionality inside vSphere (and I haven't progressed to vCenter yet, though I plan to do so). I was hoping to get LSI MSM installed, but I am having an error.

I installed MSM 8.00-05 from LSI's website and used the vmware install script. I opened the ports needed in ESX's firewall. I loaded the MSM client on either a guest on the ESX host or on my own local machine and I am able to connect and login to the server.

Everything loads and I have just enough time for a couple of clicks (about 20 seconds) before I lose connection to the server. After that, I cannot reconnect unless I go to the server console and issue

/etc/init.d/vivaldiframeworkd restart

. Then I can connect again, but only for another 20 seconds.

Due to this, I cannot manage my server, such as rebuilding a degraded array or configuring a new array, without rebooting the host, which is an unacceptable solution.

If I have an ssh connection open to the server when the framework crashes, I do get some terminal output. It is:

# *** glibc detected *** ../jre/bin/java: double free or corruption (!prev): 0x081d2d40 ***

======= Backtrace: =========

/lib/libc.so.6[0x8cd121]

/lib/libc.so.6(cfree+0x90)[0x8d0bf0]

/usr/local/MegaRAID Storage Manager/jre/lib/i386/client/libjvm.so[0x623f009]

/usr/local/MegaRAID Storage Manager/Framework/libstorelibjni.so(_ZN7JNIEnv_24ReleaseByteArrayElementsEP11_jbyteArrayPai+0x1f)[0xed4f26e7]

/usr/local/MegaRAID Storage Manager/Framework/libstorelibjni.so(Java_plugins_StorelibPlugin_processNativeCommand+0x1cb)[0xed4f1d59]

/usr/local/MegaRAID Storage Manager/jre/lib/i386/client/libjvm.so[0x621b25d]

/usr/local/MegaRAID Storage Manager/jre/lib/i386/client/libjvm.so[0x630f998]

/usr/local/MegaRAID Storage Manager/jre/lib/i386/client/libjvm.so[0x621ab70]

/usr/local/MegaRAID Storage Manager/jre/lib/i386/client/libjvm.so[0x621abfd]

/usr/local/MegaRAID Storage Manager/jre/lib/i386/client/libjvm.so[0x628b265]

/usr/local/MegaRAID Storage Manager/jre/lib/i386/client/libjvm.so[0x63a03dd]

/usr/local/MegaRAID Storage Manager/jre/lib/i386/client/libjvm.so[0x6310ac9]

/lib/libpthread.so.0[0x9b549b]

/lib/libc.so.6(clone+0x5e)[0x93533e]

======= Memory map: ========

00846000-00860000 r-xp 00000000 08:15 196046 /lib/ld-2.5.so

00860000-00861000 r-xp 00019000 08:15 196046 /lib/ld-2.5.so

00861000-00862000 rwxp 0001a000 08:15 196046 /lib/ld-2.5.so

00864000-009a2000 r-xp 00000000 08:15 196047 /lib/libc-2.5.so

009a2000-009a4000 r-xp 0013e000 08:15 196047 /lib/libc-2.5.so

009a4000-009a5000 rwxp 00140000 08:15 196047 /lib/libc-2.5.so

009a5000-009a8000 rwxp 009a5000 00:00 0

009aa000-009ac000 r-xp 00000000 08:15 196048 /lib/libdl-2.5.so

009ac000-009ad000 r-xp 00001000 08:15 196048 /lib/libdl-2.5.so

009ad000-009ae000 rwxp 00002000 08:15 196048 /lib/libdl-2.5.so

009b0000-009c3000 r-xp 00000000 08:15 196051 /lib/libpthread-2.5.so

009c3000-009c4000 r-xp 00012000 08:15 196051 /lib/libpthread-2.5.so

009c4000-009c5000 rwxp 00013000 08:15 196051 /lib/libpthread-2.5.so

009c5000-009c7000 rwxp 009c5000 00:00 0

009dd000-00a02000 r-xp 00000000 08:15 194744 /lib/libm-2.5.so

00a02000-00a03000 r-xp 00024000 08:15 194744 /lib/libm-2.5.so

00a03000-00a04000 rwxp 00025000 08:15 194744 /lib/libm-2.5.so

00a06000-00a0f000 r-xp 00000000 08:15 194796 /lib/libcrypt-2.5.so

00a0f000-00a10000 r-xp 00008000 08:15 194796 /lib/libcrypt-2.5.so

00a10000-00a11000 rwxp 00009000 08:15 194796 /lib/libcrypt-2.5.so

00a11000-00a38000 rwxp 00a11000 00:00 0

00b19000-00bf4000 r-xp 00000000 08:15 830095 /usr/lib/vmware/lib/libstdc++.so.6

00bf4000-00bf8000 r-xp 000da000 08:15 830095 /usr/lib/vmware/lib/libstdc++.so.6

00bf8000-00bf9000 rwxp 000de000 08:15 830095 /usr/lib/vmware/lib/libstdc++.so.6

00bf9000-00bff000 rwxp 00bf9000 00:00 0

00c55000-00c68000 r-xp 00000000 08:15 196064 /lib/libnsl-2.5.so

00c68000-00c69000 r-xp 00012000 08:15 196064 /lib/libnsl-2.5.so

00c69000-00c6a000 rwxp 00013000 08:15 196064 /lib/libnsl-2.5.so

00c6a000-00c6c000 rwxp 00c6a000 00:00 0

00cab000-00cb2000 r-xp 00000000 08:15 196057 /lib/librt-2.5.so

00cb2000-00cb3000 r-xp 00006000 08:15 196057 /lib/librt-2.5.so

00cb3000-00cb4000 rwxp 00007000 08:15 196057 /lib/librt-2.5.so

06000000-0642a000 r-xp 00000000 08:15 941214 /usr/local/MegaRAID Storage Manager/jre/lib/i386/client/libjvm.so

0642a000-06444000 rwxp 0042a000 08:15 941214 /usr/local/MegaRAID Storage Manager/jre/lib/i386/client/libjvm.so

06444000-06864000 rwxp 06444000 00:00 0

08048000-08052000 r-xp 00000000 08:15 941045 /usr/local/MegaRAID Storage Manager/jre/bin/java

08052000-08053000 rwxp 00009000 08:15 941045 /usr/local/MegaRAID Storage Manager/jre/bin/java

080ef000-08350000 rwxp 080ef000 00:00 0

ebc83000-ebc84000 ---p ebc83000 00:00 0

ebc84000-ec684000 rwxp ebc84000 00:00 0

ec684000-ec685000 ---p ec684000 00:00 0

ec685000-ed085000 rwxp ec685000 00:00 0

ed085000-ed094000 r-xp 00000000 08:15 196062 /lib/libresolv-2.5.so

ed094000-ed095000 r-xp 0000e000 08:15 196062 /lib/libresolv-2.5.so

ed095000-ed096000 rwxp 0000f000 08:15 196062 /lib/libresolv-2.5.so

ed096000-ed098000 rwxp ed096000 00:00 0

ed098000-ed09c000 r-xp 00000000 08:15 194726 /lib/libnss_dns-2.5.so

ed09c000-ed09d000 r-xp 00003000 08:15 194726

A google search revealed trying

export MALLOC_CHECK_=0

but that did not resolve the issue.

Can someone point me to some solution? I can't imagine that this is a new issue, so what does everyone else do when they need to rebuild an array?

Thanks!

Tags (3)
76 Replies
randysierrascie
Contributor
Contributor

That's what I did: I did a fresh install of MSM 2.91-05 on an XP32 bit vm and on a physical. In both cases, when I open MSM it sees my raid card in my physical desktop running XP64, but it does not see the card in the ESXi server, on the same subnet (we only have one here). I try to manually enter the ESXi server's IP but all that gets me is a blank slate in MSM - no machines to select. Is my installation of current MSM in my desktop somehow poisoning the network? Guess there's only one way to find out.

0 Kudos
randysierrascie
Contributor
Contributor

I uninstalled my copy of MSM (newest version) on my desktop. The old version installed on another box still doesn't see the esxi server's card. I am running ESXi Version 4.1.0 Build 260247.

0 Kudos
JaroF
Enthusiast
Enthusiast

Make sure you allowed the old version through your firewall on your desktop

0 Kudos
Mindflux
Contributor
Contributor

Yes what version of MSM did you end up using? I am experiencing the same thing with ESXi 4.1. I used to be able to manage 4.0 with MSM 6.x but now I have the same symtoms. (MSM Doesn't see server to connect to)

*Edit: I've since tried the 2.91-05 link above with the same (no luck) results.

0 Kudos
randysierrascie
Contributor
Contributor

Are there any VMware folks who could shed some light on this?  I suspect that, now, the only way to make this work is to buy Vsphere and some other stuff.  If that is the case, please just say so.

0 Kudos
stripe
Contributor
Contributor

I have the same experience with Intel RS2WC080. The card itself is performing well, shows disks states by CIM and is recognized even by name by ESX(i) 4.0, 4.1. But the software from intel is cracky.

I have post to their intel forums for help, didnt help, someone post me, that the card does not support VMware. Ehh? Even Intel has own driver for older 3.5 and 4.0 <update2 and version of MSM seems supporting the VMware. However service mrmonitord does not even start because of "error while loading shared  libraries: libxerces-c.so.28: cannot open shared object file: No such  file or directory". Tried to get missing libs into console but I fall  into dependencies hell and service vivaldiframeworkd crashes shortly because of Java - same error as above.

So I have played with your hints, earlier versions of MSM, ESX, ESXi but I did not succeed. I dont want to return the card to store because of bad Java in EWC2 8.0.5-4 in ESX. I would like to change my old server in datacenter for a new one but I am limited by dimensions of the box there so I cant simply get an OEM DELL or HP server into box.

So are there any news about this topic?

0 Kudos
randysierrascie
Contributor
Contributor

OEM Dell or HP?  Why?  Smart money goes with Supermicro.

I mean, just consider this one fact: HP and Dell do not make motherboards. 

So, what brand of board will you get in your next Dell or HP "box"?  Whatever's cheapest?  Could be, no?

Just saying.

0 Kudos
stripe
Contributor
Contributor

Te goal is not to hack the box to go smoothly with software. Id like to have a box which just works. With Dell or HP I have guarantee that it works and have support NBD. Just Id like to use the old systemI have which is my made, and fit into datacenter box, to go with ESX with supported controller and IPMI.

Even Supermicro does good boards I do not want to spend further money in pieces of hardware hoping it will work smoothly with VMware together. Now I am looking at Dell T310 so we will see.

0 Kudos
stripe
Contributor
Contributor

wrote a bugreport to Intel, this week guys from Intel responded to me that they duplicated the problem on some of their controllers and they will work on resolution after holidays.

with the wish of Merry Christmas

Stripe

0 Kudos
beerwatch
Contributor
Contributor

I run into similar problem. May I ask whether there is any progress "after holidays"? Thanks a lot.

B-

0 Kudos
stripe
Contributor
Contributor

yep as I am so far informed, the pending version RWC2_linux_8.17-06 is still tested and its in RC phase. It corrects crash issue on ESX 4.0 and 4.1 and other issues reported recently. Hopefully will be available on intel site for download during February. I am promised to be informed when it happens. So Ill drop a line here ASAP after that.

0 Kudos
briandatamatter
Contributor
Contributor

Just found this;

http://www.hkepc.com/forum/viewthread.php?tid=1574798

Is downloading now. Will give it a try and report back here.

0 Kudos
stripe
Contributor
Contributor

yep,

confirm, this is the promised version which fixes the crash problem on ESX hosts

just I will ask Intel where is its announced version.

link for LSI version is here http://www.lsi.com/storage_home/products_home/internal_raid/megaraid_sas/entry_line/megaraid_sas_924...

0 Kudos
Mindflux
Contributor
Contributor

What crash problem? I just want to be able to connect MSM to ESXi 4.1 to manage the arrays.

stripe
Contributor
Contributor

then you just need to reason LSI into make the installation VIB bundle for ESXi like Dell with their OpenManage does. Of course Id prefer ESXi way also

IMHO Id prefer APC PowerChute to be bundled to ESXi also. But this strongly depends on the vendors if they choose, not this discussion

This topic is about MSM for ESX console where needed daemons were constantly crashing

0 Kudos
Mindflux
Contributor
Contributor

well there IS an offline-bundle you can obtain from their ESXi driver disk "ISO", however it's weird that the VIB isn't included with ESXi by default since the cards are supported via the megaraid_sas driver.

I haven't tried injecting the offline-bundle into my ESXi install. Need to have a maintenance window to do so.

0 Kudos
stripe
Contributor
Contributor

if can you post me link for that bundle I can test it for you, have test box.

Anyway Ill try finally 8.17-20 version for ESX for behaving.

0 Kudos
Mindflux
Contributor
Contributor

This link will take you to a ISO for the megaraid_sas.o driver. Inside the ISO will be a bundle.

http://downloads.vmware.com/d/details/esx4_lsi_megaraid_dt/ZCV0YnRlcHBidHdw

0 Kudos
stripe
Contributor
Contributor

so I can confirm that 8.17-20 is stable on ESX 4.1, U1

Intel version of MSM will be ver. 9 I am told that will be hopefully next week.

MindFlux: I think that bundle is just driver as I looked into, its just few bytes, or did you try it earlier??

0 Kudos
Mindflux
Contributor
Contributor

VIB in bundle is supposed to let you manage card from remote MSM install.

Sent from mobile, please excuse any typos or errors.

0 Kudos