New controller, read errors only on 8TB disks


Recommended Posts

Have an M1115, added what is supposed to be an LSI 9211-8i.

 

On boot, the BiOS for the controllers shows an SAS9211-8i and an SAS9200-8e.  The 8e is the new card.

 

I immediately had hundreds of read errors, but only on my 8TB Seagates.  These were connected to the new controller.  Other disks were also connected and no read errors were present.  A parity check, as a test, was attempted.  250-400KB/s, eventually going to 0.0KB/s, with over 600 days to complete.

 

I swapped cables so the 8TB disks would run on the old controller.  Now I am getting 15-28MB/s on a parity check, so far no read errors, but obviously these speeds are atrocious.  

 

From what I can gather, the "8i" means 8 internal connections, while the "8e" means 8 external connections, ie, SAS expander.  So perhaps the new card is flashed with the incorrect firmware?

 

Diags attached.

 

Any help appreciated.

ffs2-diagnostics-20170730-0542.zip

Link to comment

Additional research.

 

Could be a counterfeit card.  This is my main concern.

 

According to what I've read online, the 8e is indeed a model number expressing external connectors, which have a cable model type different from internal connectors.  I doubt the cables are electrically different, but if they are, then if the card thinks it should be communicating over one type of cable (external) and I have it connected with internal cables, I am worried about data corruption.  

 

If the card is counterfeit, I am worried about data corruption.

 

If the card is legit and was flashed with incorrect firmware, I am worried about data corruption.

 

I have a parity check going again and it seems to have settled at 20-29MB/s..... About 25% of what I usually get.  So something is very definitely wrong somewhere down the line.

 

While I do not appear to be getting read errors now, I am getting thousands of these:

 

Jul 30 06:46:06 ffs2 kernel: scsi target1:0:4: handle(0x000c), sas_address(0x4433221105000000), phy(5)
Jul 30 06:46:06 ffs2 kernel: scsi target1:0:4: enclosure_logical_id(0x500605b002c8b32b), slot(5)
Jul 30 06:46:06 ffs2 kernel: sd 1:0:4:0: task abort: SUCCESS scmd(ffff8802c42b1c80)
Jul 30 06:46:08 ffs2 kernel: mpt2sas_cm0: log_info(0x31110630): originator(PL), code(0x11), sub_code(0x0630)
Jul 30 06:46:39 ffs2 kernel: sd 1:0:2:0: attempting task abort! scmd(ffff88030c4d2c00)
Jul 30 06:46:39 ffs2 kernel: sd 1:0:2:0: [sdd] tag#4 CDB: opcode=0x88 88 00 00 00 00 00 02 4d b6 50 00 00 04 00 00 00
Jul 30 06:46:39 ffs2 kernel: scsi target1:0:2: handle(0x000d), sas_address(0x4433221103000000), phy(3)
Jul 30 06:46:39 ffs2 kernel: scsi target1:0:2: enclosure_logical_id(0x500605b002c8b32b), slot(3)
Jul 30 06:46:39 ffs2 kernel: sd 1:0:2:0: task abort: SUCCESS scmd(ffff88030c4d2c00)
 

Other posts on this forum point to a controller problem.  I am re-using old cables, but they were used (successfully) with another controller in another system.

 

Perhaps the controller is flat out bad?

 

I really want to just power down the server until this is settled, but there is data there that would be nice to access.  I can live with slow speeds, for now, but as I said, my chief concern of course is data corruption.  Is it safe to proceed or should I just power down the server until I have time to swap out the new controller and return to the old config?

 

 

Link to comment

So an 8i will have 2 internal SFF8087 ports, and the 8E will have 2 external SAS connectors on the PCI Bracket

Can you try remove the new card physically from the server? It may be that the two cards are too far apart in firmware to be playing nice in the one setup.

I run my LSI cards on P20 across the board, some people prefer P14 as well.

If you can try removing the new card, and for curiosity, post a photo of the cards too

Cheers


Sent from my iPhone using Tapatalk

Link to comment

As suspected, the card thinks its an 8E, but its not.  Might that be an issue?

 

Back to original config, old LSI purring along, onboard SATA controlling the rest.  140MB/s parity check, will eventually settle down to 80-90MB/s as usual.  So we're back in business.  Old card is out, gonna mess with firmware and cables on my next days off.  Otherwise its going back to the fleabay user I got it from.

 

Pics attached.  I realize I cut off the firmware revision, but yeah, the firmware is different.

 

Old card is on bottom (vertical sata ports), new card is on top (ports on the end of the card).  

 

IMG_20170730_072838.jpg

IMG_20170730_045357.jpg

IMG_20170730_045330.jpg

Link to comment

Just two questions, then. 

 

Virtually every single LSI I've seen online has had "LSI" silk screened to the left of the "PCI Express" screening.  Mine doesn't have LSI printed there.  It was sold as a "bulk" item.  I have no reason to believe LSI/Broadcomm would NOT do OEM/bulk stuff, but just wanting to confirm, if anyone knows.... Will *all* LSI mfg'd products have LSI marked on the board?

 

Finally, does the "9200-8e" model number expressed by the BIOS call for any concern, or is this simply a text string that can be manipulated arbitrarily and without any effect on the function of the card?

 

Thanks again all.

Link to comment

Confirmed.  Seller fzecn on fleabay is selling counterfeit LSI cards.  In his very broken English he said it was a "third party unit."   I started a return process and, despite the original listing showing a US seller, the return to address is some 50 word address in China.  I would say I'm shocked, but I'd be lying.  So steer clear of this seller.

 

--

 

 

 

Dear sir:
It is thred party unit, It is may be don't support Avago drive.
Best regards.
 
 
 

Your previous message

alongside my existing 9211 it shows up differently!!! is this a counterfeit card? or just flashed wrong???

fzecn:

Dear sir:
It is not 9200-8E, it is 9211-8I, please try to install it with the correct drive.
Best regards.

Your previous message

This controller is showing up as a 9200-8e. I believe this means it thinks it has 8 external ports, ie, an SAS expander. I have it running alongside a 9211-8i and the 9200-8e is causing thousands of disk read errors. I believe this was incorrectly flashed to the wrong firmware. Can you advise please?
Link to comment

Just to clear some misconceptions on @tucansam

None of the LSI cards have a SAS expander. all of the cards have enough chips and PCIe Lanes to properly drive all the interfaces.

An external SAS port (SFF-8088) is electrically the same as an internal port (SFF-8087), the only difference being the shielding requirements and the rigidity of the connector. There are adapters for converting back and forth passively.

 

That said. looks like you have a rotten lemon :(

 

Link to comment

Thanks, I appreciate the clarification.  The more I look into this seller, the more shady he becomes.  I should have known better, but I let the US address fool me, and it was sort of a panic buy since my array was hosed and I needed a good card ASAP.  Yet another lesson learned on that damned auction site.  

Link to comment

Update for anyone interested. 

 

I sent the serial number of the card to Broadcomm, who has no record of said serial number, and thus concluded, as did I, that the card is a fake. 

 

In addition, I now have the card working, in another system, free of errors.  However, I am not able to exceed 21MB/s during any disk operation. 

 

The seller refunded my money after calling him out, and asked that I send the card back to him for testing (to a China address), guaranteeing he would refund my money spent on shipping.  The card initially shipped from a US address, and the seller's listing shows a US address, thus clicking "US Only" in the fleabay filter shows his listings (despite him being a Chinese seller selling Chinese knock-offs).  Misleading.  After telling him that I would not be shipping the card to China, and requesting a US address, he told me to "keep or throw away the card."  So its going in the trash.

 

Thanks again to all for the help.  I limit my purchases on fleabay these days as much as possible, and will be even more discerning going forward.

Link to comment
23 minutes ago, tucansam said:

 

I sent the serial number of the card to Broadcomm, who has no record of said serial number, and thus concluded, as did I, that the card is a fake.

 

Yep, since they screwed I don't use ebay anymore but when I did for buying my LSI controllers I always preferred to buy an used server pull even if there were new controllers from china or even the US for the same price, when the price looks too good to be true there's often a reason.

Link to comment

So it sounds like the 9201-8i is working fine. But the 9200-8e is not.

 

This is the second problematic 9200 card reported here (other was a -16e). I would steer clear of them. They are an older generation (not sure why anyone in China would want to clone them, versus a more modern 9201 or 9211).

 

Knock off or real, if they work I am not so terribly concerned. All of my LSI cards have looked like other LSI cards I own, with the LSI stickers and branding.  But I am not an expert. I bought a new 9201-16i from Hong Kong and a 9201-16e used, and both are working well. I guess you take your chances with eBay sellers, but overwhelmingly the purchases I've seen here have been successful and the products received have functioned appropriately. (Excepting the 9200s.)

 

I know its a hassle, but if they're refunding the money, sounds like they are making it right. I would ask for a prepaid shipping label though. Payments that go outside eBay are not covered by eBay's refund policy.

 

I will still turn to eBay to buy used controllers, and just have my "caveat emptor" meter on (which sounds exactly like what happened with @tucansam here).

Link to comment
23 minutes ago, bjp999 said:

This is the second problematic 9200 card reported here (other was a -16e). I would steer clear of them. They are an older generation (not sure why anyone in China would want to clone them, versus a more modern 9201 or 9211).

 

It's not a 9200-8e, he bought a 9211-8i and besides being a fake it comes with a 9200-8e bios, I have a genuine 9200-8e and it works perfectly, it uses the same SAS2008 chipset as the 9201, 9210, 9211, etc.

Link to comment
1 hour ago, johnnie.black said:

 

It's not a 9200-8e, he bought a 9211-8i and besides being a fake it comes with a 9200-8e bios, I have a genuine 9200-8e and it works perfectly, it uses the same SAS2008 chipset as the 9201, 9210, 9211, etc.

 

 

Yes, exactly correct.  Marked as a 9211-8i, sold as a 9211-8i, pictured as a 9211-8i, the card that arrived looks like every other 9211-8i, however it boots as a "9200-8e" and, lo-and-behold, is a fake according to Broadcomm.  It functions in the technical sense, but will not sustain more than 21MB/s.

 

I also determined that the PCIe x16 slot on my motherboard is bad, and that was the cause for the read errors on any controller installed in that slot.  Any controller/cable/disk combo in that slot resulted in either no drives being detected, or read errors on the drives that were.

 

So now I am looking at a 9201-16i (as mentioned above, server pull from a US seller with 60k+ positive feedback, I think I'm safe) to replace the 9211-8i as I need more port density.

Link to comment

Got it.

 

I have a 9201-16i. Works great. I think I see the one you were looking at . ~$160. That is a very good price for that controller. And looks like a real US seller.

 

You need a PCIe 2.0 or 3.0 slot. And make sure your slot is wired for x8 or x16. Sometimes a slot has a longer connector, but is only wired for x4. Anything less and you are liable to have performance issues on parity checks at least.

Link to comment

Thanks for the reply, its nice to confirm that the 9201-16i works, I knows its listed in the compatible hardware list but hearing from an actual user is good.  

 

Yeah, it started at $179, and now, with less than 24hrs to go, its at $161.  I'm hoping in the last few hours of the listing it drops even more.  I've spent enough on various things troubleshooting this system in the last month, getting a deal on this card would be sweet.

 

I only have an x4 slot for it, until I figure out how to repair the x16 (I have to pull the MB and look for bent pins).  Mind if I ask what kind of speeds you're getting with yours?

Link to comment
5 minutes ago, tucansam said:

Thanks for the reply, its nice to confirm that the 9201-16i works, I knows its listed in the compatible hardware list but hearing from an actual user is good.  

 

Yeah, it started at $179, and now, with less than 24hrs to go, its at $161.  I'm hoping in the last few hours of the listing it drops even more.  I've spent enough on various things troubleshooting this system in the last month, getting a deal on this card would be sweet.

 

I only have an x4 slot for it, until I figure out how to repair the x16 (I have to pull the MB and look for bent pins).  Mind if I ask what kind of speeds you're getting with yours?

 

This might help.

 

It depends on how many array disks are connected to the controller. The fewer the better the parity check performance. 

 

This might help set expectations. This analysis was done at the theoretical maximum speed, so the results are overstated by 5-10% in practical use.

 

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.