Lenovo servers and "lightpath diagnostics"

[BLinux]

I saw this "lightpath diagnostics" mentioned here:

First time i've heard of such a feature, which sounds really great for troubleshooting systems when a component fails or has errors. Does anyone else (HP, Dell, IBM, etc.) have this kind of feature?

 

[WANg]

I saw this "lightpath diagnostics" mentioned here:

First time i've heard of such a feature, which sounds really great for troubleshooting systems when a component fails or has errors. Does anyone else (HP, Dell, IBM, etc.) have this kind of feature?

Hm. It's basically a bunch of LEDs that light up telling you what components failed? What happens if the LED fail?

 

[Aestr]

Hm. It's basically a bunch of LEDs that light up telling you what components failed? What happens if the LED fail?

Then you troubleshoot as if you didn't have them. It's really just a tool to speed up detection of some likely problem components, but shouldn't be taken as the only or overriding diagnostic tool. A failure LED indicates a problem likely exists. A lack of an LED does not mean that component is not a problem. It could be a failed LED as you mention or just an issue that the light path is not capable of detecting.

I've dealt with users who took these a little too seriously and would blindly follow the panel refusing to believe a problem could exist if the component had an LED that was not lit even though there was plenty of evidence pointing them to the problematic piece of hardware.

 

[BLinux]

Hm. It's basically a bunch of LEDs that light up telling you what components failed? What happens if the LED fail?

I think so (to the 1st statement). As for the LED failing.. well, you could say that about everything else... what if the BMC firmware had a bug? what if the BMC chip failed? etc.

i think this is actually meaningful considering my recent dealings with EDAC errors on one of my servers. in that case, I think there was in fact a bug in the EDAC code reporting errors on the wrong DIMMs. but even before that, it took some effort to decipher what DIMMs the messages were pointing to; even if they were pointing to the wrong ones.

 

[WANg]

Then you troubleshoot as if you didn't have them. It's really just a tool to speed up detection of some likely problem components, but shouldn't be taken as the only or overriding diagnostic tool. A failure LED indicates a problem likely exists. A lack of an LED does not mean that component is not a problem. It could be a failed LED as you mention or just an issue that the light path is not capable of detecting.

I've dealt with users who took these a little too seriously and would blindly follow the panel refusing to believe a problem could exist if the component had an LED that was not lit even though there was plenty of evidence pointing them to the problematic piece of hardware.

So basically, it's like parsing logs on IPMI or going through SELs, just another tool in the sysadmin troubleshooting arsenal, not something 100% definitive.

 

[MiniKnight]

If that guy can be the head of Lenovo's 1U and 2U servers @Patrick could run their server division.

 

[Evan]

IBM have had the concept of LED codes on their servers for i dont’t know but maybe 30 years, they had a real display of earlier systems and today it’s shown via the management systems. It even used to show different info about OS errors and boot.

 

[WANg]

If that guy can be the head of Lenovo's 1U and 2U servers @Patrick could run their server division.

I kinda smirked when that guy acted so surprised about the pair of M.2 slots in the back mezzanine. What, never seen the paired SD card slots on the back of a 12th Gen Dell PowerEdge? And seriously, if it's just for boot media (like for bringing up ESXi), why spend the money on M.2?

 

[Evan]

Dell have had that kind of setup a while, HPE kind of have some some similar options like slots on the risers for m.2 SATA disks but can’t mirror them with controller so was kind of wandering the point unless it’s to boot a Linux OS using LVM