S5E19: "RAM it in Hardware"

Posted
Comments 0

Navigation
Previous EpisodeNext Episode
Log
Recorded (UTC) Aired (UTC) Editor
2020-10-29 02:43:30 2020-11-07 08:24:55 "Edita"
Verification
Format SHA256 GPG Audio File
MP3 465a1464e0d3fc8a1d61853fffc9fe35936f1761d549913b9bdc4bc281a45b76 click click
OGG 0a216245a0a0974f5ddc12c5e230153dac7e7843a931ef29589d0eff9919fdca click click
Quicklisten:

In this episode, we delve deeper on the topic of hardware for servers – specifically, Jthan features a lengthy look into ECC RAM.

I apologize if my audio seems a bit off in this one; I lost my local recording so we had to rely on Mumble’s recording!

Just the Tip

  • Paden talks about how not to perform a software/platform/etc. migration.

Notes

Starts at 25m34s.

I was drinking a Guinness Draught. Paden was drinking Grey Goose vodka and orange juice. Jthan was drinking MIchelob Ultra.

  • ECC (“Error-Correcting Code”) RAM
    • ECC RAM with ZFS is a mixed bag – it has some tradeoffs.
      • ECC RAM is a bit slower.
    • “Segfaults instead of kernel panics”, for an abstract way of looking at it.
    • They also log hardware errors in the MCE (“Machine Check Exception”) log.
  • Power supplies (38m33s)
    • How many PSUs?
      • For availability, Jthan thinks you should get at least a single failover (2x total) PSU chassis/board.
    • For boxes with high power consumption, keep in mind the failover factor needed!
  • Disks (Hard drives) (42m44s)
    • Jthan recommends enterprise disks because they last longer.
    • If you’re using platter disks, remember about vibrational sensitivity (and the vibration the disks themselves generate as well)!
    • Jthan whores about ZFS again and then promptly forgets to talk about it. He really wanted to talk about error recovery control.
    • He also cautions about how spin-up/spin-down affects warranty and guaranteed lifetime of disks.

15 Clams

In this segment, Jthan shares with you a little slice of life. The title is a reference to this video. (2m16s in)

Starts at 59m10s.

Jthan tells us about the datacenter watercooled in a pond.

And by that he means make sure you have backups and a restore procedure. Also make sure you have a failover plan or procedure for your secondary/failover backup server if it fails.

Errata

  • This is the poll we were referring to in the intro. Relax, we recorded this before the election even started.
  • Dook in Peace. Charlie Buckets (the ferret). You were a good boy. :(
    • I was wrong, he was actually pretty young. It’s not unheard of for them to die around 4 years old from cancer or other medical conditions, but he was still fairly young (~7 years is roughly the upper limit if they develop cancer. “Old” would be 13 if natural causes/old age.). He died of a heart attack. He was 3.5 and had degenerative cardiomyopathy.
    • You can find photos of him here: 1, 2, 3, 4
    • Still miss ya, bud.
  • I reference the Good, the Bad and the Ugly (which Jthan hasn’t seen because he doesn’t watch good movies apparently).
  • ZFS was written originally by three people.
    • Jthan said he was thinking of Matt Ahrens, specifically this quote from him.
  • Regarding platter drives and vibrational sensitivity, this will forever be one of my favourite videos.
  • Member loken in our Discord has pointed out to us the following. Thanks, loken!

Music

Music Credits
Track Title Artist Link Copyright/License
Intro Inspired Kevin MacLeod click CC-BY 4.0
Outro Trouble With Tribals Kevin MacLeod click CC-BY 4.0
(All music is royalty-free, properly licensed for use, used under fair use, or public domain.)

Author
Categories

Comments

There are currently no comments on this article.

Comment...

Enter your comment below. Fields marked * are required. You must preview your comment before submitting it.