Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[AArdvark]

Totally agree with everything said re PSU.
I would power up using a temporary PSU out of something else you trust if you do not have a spare and repeat the Blu-ray backup to be extra sure 
The 'out of spec' comment is very true for a sata power cable to burn/melt.

I know there is a lot of argument about good/bad drives but can only comment on my experience.
There is always someone who can argue the 100% opposite view !!
I went with Toshiba as a 'punt' but also because they had recently acquired Hitachi Global Storage Technologies (HGST) assets via WD buying the assets and selling them on. 
My personal experience of HGST drives has been good so I reasoned that maybe Toshiba Drives would be OK.
I now have Eight drives purchased over a period of time via 2 different vendors and they are all working fine .(Touch wood)
Worth looking IMHO.

[sevenlayermuddle]

As Kitz pointed out, it only seems to affect the Molex to Sata adapters.  I find that a little illogical since it is the 'disk' end that burns, it really shouldn't matter whether it's sourced from an adapter, or a native power supply lead.   All I can possibly imagine is that there is just one factory turning them out, across all brand names,  and it is a dodgy factory?   

Whatever the reason, getting away from the adapters would in any case mandate a new PSU with native Sata leads.   I am beginning to like the idea of just building a new server though.  The Proliant Microserver that Rizla linked looks good value, and I can effectively discount it further by knocking off what I'd save on the PSU.   I'm just trying to find some figures for idle power consumption, but I'll bet it'd be a worthwhile improvement.

Biggest problem of all will be finding the time, procrastinating over which Linux to install, etc.  It's not just the server; to make it worthwhile I'd end up updating to latest MythTV, so  I'll also  be rebuilding the boot USBs for the two diskless Myth front ends that sit under the TVs.  It was so easy when I set all this up.  I was working 5 days a week, so did it at weekends.  Now theoretically 'retired', I seem to have no weekends any more, in which to get things done. 

[AArdvark]

Can vouch for the Microservers in general have two of them (older models).
Good solid kit that works very well.
Running VMware ESXi / Solaris / Napp-it on zfs using Adaptec 9211 HBA's (reflashed IBM ServeRAID M1015's) with lots of disks. 
Love VMware stuff, Solaris is very good but mind bending  if you have used *nix previously and zfs is 'God's Filesystem'   

You obviously had a better work/life balance than me.

It was not unusual to find your weekends getting used up for 'Urgent' work and/or travel.
Lots of promises of rewards ('in heaven') but mainly busy busy all the time.
I did really enjoy the work but the lack of 'life' does catch up after a while. 

[sevenlayermuddle]

You obviously had a better work/life balance

I always made sure of that, no matter how much I was enjoying the work.  Any weekend overtime was scrupulously claimed and usually taken back as time off in lieu.

 At the end of the day it, working for an employer is only ever a job, the rest of your life is your own.  And far more important.   

[AArdvark]

You obviously had a better work/life balance

At the end of the day, working for an employer is only ever a job, the rest of your life is your own.  And far more important.   

So true, I was slow to learn that lesson. 

[sevenlayermuddle]

The Proliant Microserver that Rizla linked looks good value

Oops, Senior moment there, I just noticed it was Ronski that linked the Proliant Micro.   

Thank you Ronski, I remain tempted.   

[AArdvark]

The Proliant Microserver that Rizla linked looks good value

Oops, Senior moment there, I just noticed it was Ronski that linked the Proliant Micro.   

Thank you Ronski, I remain tempted.   

Do not forget to look for the Cashback offers from HP.
They have been doing Cashback on Microservers for years.
One of the reasons they were\are so popular as well as being good kit overall.

[guest]

FWIW a friend in NZ has had an identical issue and the common denominator seems to be the blu-ray writer having been recently in use. Again a molex<->sata power cable is being used but its nothing to do with Amphenol according to him.

[sevenlayermuddle]

FWIW a friend in NZ has had an identical issue and the common denominator seems to be the blu-ray writer having been recently in use. Again a molex<->sata power cable is being used but its nothing to do with Amphenol according to him.

In my case, of all the factors that night have been involved, the Blu-Ray drive is off the hook.    It resides in our XP system, totally separate from the Linux server that had the burn up, connected only by Lan.   

With hindsight I suspect I maybe got some warnings as streaming  Myth TV (from that very disk) had developed an occasional habit of freezing for a moment.   That really shouldn't ever be happening (all wired LAN), only time it happened in the past was precursor to hardware failures, router and switch.

The other clue might have been, twice after power cuts recently the server has rufused to restart, just a beep or two then powered itself off again.  It's normally headless so I have to hoik it into the other room and connect up a monitor to diagnose boot issues, at which point it started perfectly both times.    Leaving me none the wiser.

And getting into the realms of fantasy, but believable, who knows... a voltage spike associated with these power cuts may have played a part too?

[sevenlayermuddle]

Thinking more on this, as I will want to resurrect the old system regardless of whether I replace it, so I need a plan of attack that doesn't involve fire extinguishers.   There does seem to be a large body of evidence, from Internet posts, that the problem is specific to Molex/Sata splitters.   The more I think about it, that is simply nonsensical, if the Molex was overloaded the fire would occur at the Molex, not at the Sata end.

But I have a theory... The Molex/Sata splitters generally seem to have quite short, stubby, leads.   I have found in the past that when plumbing in Sata disks, the stiffness and weight of cable attached has pulled out a Sata cable or two, and the short stubby splitters may well make that problem much worse.   I reckon a connector that was just and no more making contact, but not quite fully home, could easily find enough resistance to get hot?

What chance then, that this is behind the problem... that the fires are simply caused simply by foolish folks (self included) failing to make a proper final check on all connectors before putting the lid back on the  PC?

[roseway]

That sounds very plausible. I've had plug and socket connections get hot for that reason more than once, and pushing them firmly home has averted a fire.

[Weaver]

> I reckon a connector that was just and no more making contact, but not quite fully home, could easily find enough resistance to get hot?

I agree, sounds very plausible.

[sevenlayermuddle]

Having returned from my 'travels', I've  finally been able to examine these connectors again and have a fairly convincing explanation....

On close inspection, the undamaged limb of the Sata splitter has tiny fracture marks both corners of the plastic shell of the Sata connector itself, which I reckon will greatly reduce the clamping pressure against the connectors.  Hence, assuming the burned limb had similar fractures before the burn up, it probably was a poor connection that caused the initial overheating.  But caused by weak clamping rather than being my own fault for 'leaving them loose'.

Moreover,  there are also a few directly connected  PSU Sata leads, and none of them show the fractures.  And then I noticed, the shells on these connectors look to be made of thicker plastic.   I've some vernier calipers and although a bit big and clumsy for the job they do seem to confirm, the 'good' Sata shells are made of plastic about 0.8mm thick, the dodgy ones that have split are just 0.5 mm. 

This is good as, providing I am able to confine myself to connectors made of the thicker plastic, I shall regain a bit of trust in the machine.    But I have also now ordered a new Proliant Microserver, which I shall trust even more.   

[sevenlayermuddle]

If the SATA cable connector melted then the PSU is going to have future issues IMHO as the current taken to do that is going to be well out of spec.

And you weren't wrong.   

As mentioned elsewhere, I now have a new server.   But I have been using the old server for the purposes of data migration from the affected disk, as I don't want to risk 'contaminating' the new server's integral Sata backplane by installing a disk with dodgy connectors.

The old server does work, but it is very flakey.   Three times out of four when I apply power, the PSU just repeatedly cycles and trips.   The fourth time, if I am lucky, it comes up and stays up.  It also seems a bit confused about the system battery, keeps telling me it is flat, even though it's quite new and is not flat (I have even tried a brand new one).   I am now running it without a battery at all, manually reloading default BIOS settings and date on every boot.

This is not a system I would want to leave unattended...   

[NewtronStar]

The old server does work, but it is very flakey.   Three times out of four when I apply power, the PSU just repeatedly cycles and trips.
This is not a system I would want to leave unattended...   

The only way to sure is to install a new or known to work PSU into old server and if it's still flakey then the suspect could be one of many power regulators on the PC motherboard was damaged.