Advice about 7100 SSD issue

markgca

I have a Netgate 7100 1u that has been running for almost 7 years. It came stock with a 32GB eMMC and i ordered it with the upgraded M.2 SATA SSD installed.

Recently checked the diagnostic data; SMART still shows good, but the wear indicator is at 108%, not good. This is on the SSD as i haven’t used the eMMC at all.

I plan on upgrading the unit in a year or two to a 8200 or equivalent, but there is no hurry because this one just works and is fast enough, so in the meantime…

Replacing the SSD isn’t a big deal, just the time to remove it from the rack, take the motherboard out, replacing it, etc. I’m thinking a few hours total. I would rather it not be down that long.
But if i did that i would replace it with a WD Blue SSD that costs about $71 on amazon, which should be better than the BiWIN SSD that came from the Netgate factory. That should give the SSD another 7 years, more than enough time before i have to replace the entire unit anyway.

Another option is to leave the unit in place, and simply restore the config into the eMMC. i know they wear out much faster, but it has never been used, and i figure it only needs another year, so…seems like a ‘reasonable’ option. But i read horror stories about eMMC lasting a very short time before failure, and i don’t need that.

But i also saw they have a new option of converting the eMMC to pSLC mode which should cure the wear issue, or at least make it a lot better. What i don’t know is do i lose 50% or 75% of the capacity? i.e. i don’t know how many levels are stored in this version, and i understand they changed it mid run (and i don’t want to open the board up to find out, since that would negate the benefit of restoring in place). if 50% that should be ok, but if 75% probably not good since 8GB is likely not enough. i guess i could try it and find out, but then the unit is down for maybe 30 minutes minimum? doable i guess

anyway, some questions/thoughts:

any way to tell which eMMC i have (without opening the box) to see the size loss that would come with the pSLC mode?
is this pSLC mode really a cure to the inherent issue of eMMC wear?
my gut tells me to bite the bullet and replace the M.2 SSD, even though its a pain. your thoughts? is the eMMC good enough, especially with the pSLC mode?

looking for guidance on this. if it wasn’t for the down time its a no brainer; $71 isn’t much to make sure this doesn’t go down (due to the SSD), but the tradeoff restoring to the eMMC so it only goes down for an hour or less is tempting.

there is also another option; wait till it fails. but that never turns out good, so I’m not considering that strategy.

thanks for any thoughts

mark

SteveITS

@markgca It all depends on how much you write to it. see:

Most routers we set up are on ~10% eMMC wear after many years.

I don't have an answer for your hardware questions, sorry.

marcosm

The eMMC wear was improved some versions ago so the pSLC mode isn't necessary. IIRC it's somewhere around a 75% hit in capacity. If you don't need the extra storage space then I would recommend enabling it - it'd likely last longer than the SSD.

markgca

@marcosm perfect, thank you

i'll restore the config to the eMMC and we should be good until the next upgrade!

marcosm

Something to keep in mind is that after enough wear it won't be possible to enable pSLC mode. With the exception of some packages 8GB is enough for general use and if needed in the future then there's always the option of expanding with an SSD.

markgca

@marcosm
makes sense

According to df -h, I'm using about 7G now, but then there is overhead etc. and it starts out at about 30G, so 25% of that is only 8g; probably not enough, and i need some space for boot environments, which was just set to zero after the recent install.

so i think the best course of action is just bite the bullet, replace the SSD and reinstall again.

thanks for the info/help

mark

BREVISSS

Honestly Mark, if the SSD is already showing 108 percent wear, my gut says replace it and be done with it. Seven years of service is actually pretty respectable, and while SMART still says it's healthy, those wear numbers are basically telling you it's living on borrowed time.

The eMMC option is tempting because of the shorter downtime, but personally I would view it more as a temporary fallback than a long term solution. Even with pSLC mode, you're still relying on eMMC, and while pSLC can improve endurance quite a bit, it doesn't magically turn it into an enterprise SSD. Capacity reduction depends on the underlying flash type, which is why it's hard to predict exactly without knowing the specific eMMC model.

If you're planning to keep the 7100 for another year or two, I'd rather spend a few hours replacing the M.2 now than risk dealing with an unexpected failure later. The WD Blue route sounds perfectly reasonable and would probably outlast the remaining life of the appliance anyway.

You could also check "Etech Devices" for SSD options and storage upgrades, and communities like StorageReliability Hub often have good discussions around SSD wear indicators, eMMC longevity, and Netgate hardware maintenance.

markgca

@BREVISSS that post is over a month long, and two ssd's and emmc later.....

so i ended up ordering a ssd from western digital, and it failed with a day or two; you could actually watch the wear indicator climb on its own, in ONE day, to over 40%. so that went back. WD said they werent design for that type of use, but come on, two days?

then i got on ebay and purchased another ssd, same model as what came with the unit, just 128gb vs 256 and pulled from presumably a laptop. that worked for about three weeks and died

so then i installed it on the emmc and used that for about a week until i replaced it with the ebay replacement for the one that died.

and its still working (but only a couple days later so who knows)

i can replace it and reprogram it in about 20 minutes. kinda like when i was a kid and could change the clutch out on my car in under 30 minutes (when we were racing). do anything enough times and you discover shortcuts (in this case it involved not removing ANY of the wires, and just moving the board to where you could get at the ssd and swap it out).

anyway, we are sitting at 0% wear so thats a good sign, hopefully it will last a year or so.