HP / Aruba 2920 - Chronic PSU failures

[rawr]

Hi,

We've been replacing our Cisco edges switches (2950, 2960 and even some older devices) with HPE / Aruba 2920 switches for nearly 3 years. now.

Since then, I've had 4 power supply blow up, the 4th, just righ now. Has anyone experienced such a high failure rate? I have more or less 30-40  HP switches which isn't much - all 2920. Mostly 48 ports PoE+ 370W, some 24 ports. 

Meanwhile, I have a bunch more Cisco switches ranging form 5 years old to 14 years old devices running in dusty environnements (labs, storages..) which sitll hold strong, never breaking a sweat. 

I understand that HPE 2920 are lower end switches but I never would have expected to have so many failures. Is that normal and to be expected from the 2920 line? 

We went with 2920 purely because of the pricepoint but I cannot recommend that switch anymore. Are the 3810 or 2930M line any more reliable? Is it the same power supplies in these switches? 

Any input would be appreciated. I for one don't care much about Cisco vs HP but when it affects the business' productivity, it's another story. 

Thanks!

[parnassus]

So, in summary, less than forty Aruba 2920, mostly of them 48 ports PoE+, remaining 24 ports...4 PSUs hardware failures in nearly 3 years (so the average is about 1 PSU failure on 40 units per Year, at least just after the very first year of duty): Have you found any relationship between PSU failure and, I'm trying to guess here, Switch environment operating conditions (Dust/Tempersture), PoE load types and/or software versions?
Don't know Aruba 2920 (PoE+) declared MTBF but I believe that 3 years is a very low and disappointing value.

[rawr]

Hi,

The main relationship is the software version – our production software is WB.15.12.0015 which all failed switches were running. We just began rolling out WB.16.03.004.

We have a lot of very dusty environment although 3 out of 4 PSU failures occurred in our clean / controlled hub rooms.

• 1 failure occurred in a fairly dusty environment - J9729A switch with a single PoE device (Meraki)

3 out of 4 failures occurred on stack members (1x 2 members stack and 2x 3 members stacks) in clean environnements

• All J9729A devices
• 2 of which were running ~40 PoE devices (Mostly Cisco 7841 phones, a few meraki Aps)
• 1 of which was running ~3 PoE devices (Meraki APs)

In all cases, failures occurred on J9729A devices running WB.15.12.0015.

I checked our inventory and we have exactly 26x J9729A  (few others are J9728A (2920-48G) and J9727A (2920-24G-PoE+) )

So narrowing it down to J9729A devices, that’s ~1.33 failures per 26 units per year (We bought our first J9729A switches in ~nov/dev 2014). And that’s not counting the fact that we have been rolling out progressively, which would probably make the failure rate closer or above 2 failures per year per 26 units. The last failure occured on a switch deployed less than 5 months ago.

If all our 48 ports PoE switches were 2920 (we probably have ~30 more to replaces in the next few years), we’re looking at ~4 failures per year.. the fact is that we have mostly small offices, ~45 across Canada and Europe, we can’t just have 40 extra power supplies laying around for the random occasion.

I'll see if we can get more lucky with a newer firmware but I guess that I cannot in good conscience recommand that model anymore. HPE's support did not really care when I last brought it up anyway.

[parnassus]

---

rawr wrote:Hi, The main relationship is the software version – our production software is WB.15.12.0015 which all failed switches were running. We just began rolling out WB.16.03.004.

So basically PSU failures (all related to HPE X332 575W - 370W for PoE - 100-240VAC to 54VDC Modular Power Supply module (J9738A) used into Aruba 2920-48G-PoE+ Switch (J9729A)) happened on Power Supply replaceable hardware module that is older than (and not necessarily in service since) March 2014 [*] and, concurrently, on Switches that were running all that old software version until Power Supply premature departure.

Would be interesting to understand if there is a proven relationship with Software Versions...but an Hardware failure is, primarily, an hardware failure...even if the Power Supply replaceable hardware module has intelligence to communicate with the backplane (so Software matters).

Defect Search (try it on HPE My Networking portal) provides no references about either X332/J9738A or Aruba 2920 Power Supply failures...so this means that such issues, even if reported on HPE ticketing system, weren't enough to create a case history (relevant, read: that impacted a lot of customers with many frequent specific - replaceable - hardware related issues). Sigh.

I don't work for HPE so I'm just wrote my two cents here.

[*] March 2014 is the month during which your production software WB.15.12.0015 version was released, I say so supposing nobody did a downgrade to that old software version on newer Hardware.