I have 4 controllers on Proliant DL580 G7: P410i, LSI9261, P812, P410.After adding last two or one OS (Windows Server) started to crash. Eliminating last two helped (disabling them in RBSU). But I need them: I have 6 cages (each with 4 HDDs) and enclosure D2700 (20 HDDs).P812 see only D2700 via external port - therefore I can't connect cages to P812 and use additional controllers. Each handles 2X4 HDDs via internal controller port. All of controllers share the same IRQ - 10 and trying to change IRQ of one controller automatically changes others to new IRQ (in RBSU). P812 and P410 has ROM v 6.62 and P410i has v 6.40. How to fix? Server crashes even when most of the disks are disconnected from controllers - the presence of 4 controllers is sufficient for crash. (I applied updates recommended by latest SPP for G7)
-
In addition to very good answer down there, I want to mention: don't be fooled by the "same IRQ". Check which IRQs are used when target OS is booted, not in the BIOS. The interrupt handling currently is way more advanced than in those old days when one wanted to configure some IRQs by hand. – Nikita Kipriyanov Jun 10 '21 at 12:09
2 Answers
I have used lsi controllers ,
First of all:
Any brand of controllers usually try to connect each other and see each other for various configurations. And when their rom software version changes they mostly cannot recover arrays or volumes.
Second:
It is possible to force bios of other brands and different versions to different cards since cards of similar years use same LSI chip. It is called cross flashing I guess. From brand to brand from IR mode to IT mode (SATA forward only disables ram and raid on card).
What I would do is simple. If you cant figure out reason of crash from event viewer, do the following, if one bullet point fails to solve the issue switch to next
check if PSU is sufficient
switch and check the pci-e ports (maybe some metal scrap or paper in it)
examine cards visually for burn marks or broken smd's
match the rom software versions and figure out problematic one
cross flash problematic one to match all brands and put same rom software ver.
cross flash all into IT mode do your raid thingies with mdadm software raid.
Usually in these situations we buy some new cards. But sata - software raid approach is very solid. I use that on every setup. You just require to use some bash mdadm commands correctly. Its a couple of commands very simple compared to LSI manual , thousands of raid controller commands , patrol reads, consistency check planning etc.
My favorite setup is software raid 10 mdadm with bcache from multiple ssd's. It works great with iscsi and samba. You just need to adjust raid chunk and fs cluster sizes correctly.
Do it carefully not to lose data. It is the method I use but I am taking no responsibility to any damage to your equipment or data loss or any other type of damage. These cross flashing things and raid setups are risky (always and ever).
- 536
- 4
- 12
Answer above helped me to identify the problem - this is good algorithm. Sometimes we suspect something that seems more obvious - in my case usage of several RAID controllers from different vendors seemed suspicious. I checked everything from accepted answer - but the server was crashing. At the very beginning I notices warnings in event viewer - WHEA-Logger, "A corrected hardware error has occurred.". According to this article this is only warning. But eventually I switched RBSU from C3 State to "performance" and now server is stable. I hope this will help owners of HP DL580 G7. Also I added more powerful PSU.
- 13
- 5