User Matthew Ife | chat.stackexchange.com

Matthew Ife

Dec 14, 2021 10:57

So, just recently was reminded this place exists! Saw the post about Michael Hampton (very sad :( )and I've got fond memories of him and this place.

Matthew Ife

Nov 4, 2014 19:34

@Iain so, what do i do again with that slack stuff?

Matthew Ife

Nov 4, 2014 09:53

Later though :). Apparently this inconvenient stuff called 'work' is obstructing me currently.

Matthew Ife

Nov 4, 2014 09:53

OK, will do.

Matthew Ife

Nov 4, 2014 09:48

Still am.

Matthew Ife

Nov 4, 2014 09:48

I've been busy!

Matthew Ife

Sep 12, 2014 20:54

@Wesley I think i'll pass ;)

Matthew Ife

Sep 12, 2014 20:06

@ChrisS I've not really looked massively into how it generates them, they are apparently cryptographically secure though.

Matthew Ife

Sep 12, 2014 20:03

which means you get many more hits in :alnum: than with /dev/urandom

Matthew Ife

Sep 12, 2014 20:03

Still always use openssl rand. Plus you can make it spit out base64

Matthew Ife

Sep 12, 2014 20:01

If i want a large amount of bytes, I'd rather kill %user than %sys.

Matthew Ife

Sep 12, 2014 20:00

I always use openssl rand rather than /dev/urandom

Matthew Ife

Sep 10, 2014 18:54

@MichaelHampton regarding serverfault.com/q/622813/75118 you can do it - but you need CAP_SYS_ADMIN and a new enough distro.

Matthew Ife

Sep 9, 2014 20:11

iptables in each layer. I'm extra pow 15 secure.

Matthew Ife

Sep 9, 2014 20:10

@MichaelHampton I run my services in 15 layers deep network namespaces for 'extra' security.

Matthew Ife

Sep 8, 2014 01:26

@MichaelHampton I answered it because I need to sleep and cant be arsed waiting for him to get back.

Matthew Ife

Sep 8, 2014 01:12

If the data hasn't been stuffed over with another mkfs or something.

Matthew Ife

Sep 8, 2014 01:12

But its possible to remount using an alternative superblock.

Matthew Ife

Sep 8, 2014 01:12

I'm waiting for more details ;)

Matthew Ife

Sep 8, 2014 01:11

This guy should try mounting with an alternative superblock.

Matthew Ife

Sep 8, 2014 01:11

1

Q: How to recover data from a messed up drive (LVM written on top of Ext4)?

The previous administrator of a server that is now under my supervision made a mistake. He accidentally created a LVM volume (no more than pvcreate, I think, though not sure) on a disk that actually contained an Ext4 partition with data. What tools exist to recover data from a mistake like this? ...

Matthew Ife

Sep 8, 2014 01:07

So a lot of analysis of news feeds for upcoming events.

Matthew Ife

Sep 8, 2014 01:06

From what I can gather, -- not that I know much -- most of our models work on the basis of mid to long term market analysis, not subsecond trends and using peripheral data sources outside of what the exchanges provide to make ordering decisions.

Matthew Ife

Sep 8, 2014 01:03

That has had limited success in some revisions due to silly kernel bugs though.

Matthew Ife

Sep 8, 2014 01:03

We've done stuff like GSO/TSO etc.

Matthew Ife

Sep 8, 2014 01:03

Well, latency isn't as an important factor for our models (well, we want sub-second obviously but 60-70 microseconds isn't too much of a problem for us)

Matthew Ife

Sep 8, 2014 01:02

Basically if you're investing $1500000 a year on people to write you the algos for you're infrastructure, you're a moron to not pay $20000 for the fpgas and leverage your development teams to use them (that is, if you're in serious need of the latency)

Matthew Ife

Sep 8, 2014 01:01

I think having to do a userspace protocol stack may limit its implementation in anything serious.

Matthew Ife

Sep 8, 2014 01:00

If he has the time and energy, sure.

Matthew Ife

Sep 8, 2014 00:55

If the idea is to beat someone else competitively, since you're not doing the lowest possible common denominator - its not got a great degree of confidence that it will work.

Matthew Ife

Sep 7, 2014 23:34

its not a 'drop in' replacement in totality.

Matthew Ife

Sep 7, 2014 23:34

from what I gather you need to use user space protocol stacks to work with it

Matthew Ife

Sep 7, 2014 23:32

nope

Matthew Ife

Sep 7, 2014 18:39

I'm familiar with ipsec and strongswan.

Matthew Ife

Sep 7, 2014 18:34

Doesn't using 802.3ad actually require all endpoints to know whats going on?

Matthew Ife

Sep 7, 2014 15:14

Backups in the more 'true offsite' case are a huge freaking black hole of a sysadmins time.

Matthew Ife

Sep 7, 2014 15:14

Also, when you go portable media you need a means to track each tape in case of recovery. So that involves usually some other sort of backup software to keep some form of metadata tracking on it all.

Matthew Ife

Sep 7, 2014 15:12

We use to have to store some of our tapes in tamper proof tape cases, ultimately stored in a vault somehwere.

Matthew Ife

Sep 7, 2014 15:12

Trouble with tapes is how often you rotate them and they genuinely require somebody there to do tape rotations (depending on their value anyhow).

Matthew Ife

Sep 7, 2014 15:11

I was pwned by a wild newline. Dont rush me!

Matthew Ife

Sep 7, 2014 15:10

I would (personally) define 'offsite' to mean. 1. Data is stored somewhere not on the same site. And ideally 2. Data is stored on portable media.

Matthew Ife

Sep 7, 2014 15:05

I wanted to know the secret behind these amazingly fast backups.

Matthew Ife

Sep 7, 2014 15:05

I dont care. But I do file leve backups and they are woefully slow.

Matthew Ife

Sep 7, 2014 15:04

Oh, not really file level from a backup perspective then.

Matthew Ife

Sep 6, 2014 23:29

Ah, of course, kill -SUBSPACE 1 should do the trick.

Matthew Ife

Sep 6, 2014 23:26

Dont be confusing me with your techno rabble!

Matthew Ife

Sep 6, 2014 23:25

Yeah, you say that now @MichaelHampton but what about the proxy?!

Matthew Ife

Sep 6, 2014 23:20

In fact, can we make this a close reason?

Matthew Ife

Sep 6, 2014 23:20

►

Matthew Ife

Sep 6, 2014 23:18

@ewwhite heh, does it melt the brains of the peons?

Matthew Ife

The Comms Room