If certain tech companies produced carpentry tools...
-
@dhromed said:
Depends. Kernel? Reboot. Windows update.. maybe. Plain ol' software? OF COURSE NOT.
Eh, half the systems I designed you'd reboot for a config change or software change. Because the machines were imaged and it was better to have them reboot off the new image rather than trying to work up a way to patch running machines (plus it lets you test the boot sequence for all of the machines.) But obviously that's kind of a specialized setup.
@dhromed said:
HydrogenAudio?
High availability. Multiple servers behind load balancers, hot spare load balancers, database replication, file replication.. Most of the time when you see servers with high uptime it's because there are no HA measures in place so you can't take down that server without taking down services. But that's obviously a brittle architecture and when something does fail, it's going to suck.
-
@morbiuswilters said:
no HA measures in place so you can't take down that server without taking down services. But that's obviously a brittle architecture and when something does fail, it's going to suck.
Are we at the point yet were simple HA meaures are basically free? I mean, I don't manage the servers where I work, but even I see that virtualization make a lot of maintenance shit really trivial, so you skip the step of "upgrading from a hard server to VMs" immediately and having "no HA measures in place" seems ludicrously incompetent.
I'm not sure if I made myself clear.
-
@morbiuswilters said:
@dtech said: Kernel updates are updates, dumbass. [...] Did you post just to display your ignorance or did you have a point?
Could you please not insult people all the time? Yes I know you disagree with what he said, and yes you do have a point, but that's quite annoying and unnecessary, and it actually makes these forums a worse place for everyone.
-
@da Doctah said:
or even things that don't have a K in them at all, like potassium
Kalium. Like Tungsten (wolfram), Lead (plumbum), Tin (stannum) - there are reasons for them. Much like the names given to Linux packages/software. Though the latter tend to be less obvious (or non-existant)... @da Doctah said:thousand
Kilo...@da Doctah said:absolute temperature
Kelvin.
I don't have previous knowledge of the others you listed, so will plead ignorance on them.
-
@spamcourt said:
Could you stop whining? It's annoying and disrupts my viewing pleasure.@morbiuswilters said:
@dtech said: Kernel updates are
updates, dumbass. [...] Did you post just to display your ignorance or did you
have a point?Could you please not insult people all the time?
-
@morbiuswilters said:
Eh, half the systems I designed you'd reboot for a config change or software change.
Do you actually mean reboot, or do you mean restart/reload that service? I'm used to the latter in the *nix world.
-
@ASheridan said:
Windows is getting a lot better with the number of restarts it needs, but there are still moments of WTF. Any Adobe/MS Office product gets installed, we need to restart. They only happen to be some of the most popular software makers for the most popular software out there. True, about 70% of the time the restart can be ignored. A simple software install shouldn't really need the whole machine to be restarted.
Checkout PendMoves to get a list of the files that will be upated on the reboot. If there is nothing there, you probably don't really need to reboot or the reboot is required because of really bad design.
-
@dhromed said:
Are we at the point yet were simple HA meaures are basically free?
Some are.
LVM on disk arrays (in combination with hot-swap architecture) promote storage availability. The hardware itself isn't free, but prices have dropped considerably such that there's little cost difference between system with and without hotspare capability, making HA appear to be a freely-included option.
A number of years ago I rented a Linode (virtual server) which ran on hardware that developed a fault, so the provider moved all virtual servers were across the network to another host whilst still being up and running, then moved them back once the hardware had been fixed. I'm guessing the provider had to pay for some feature that permitted it, but to customers it meant no loss of service whilst the underlying hardware was being maintained.
-
@dhromed said:
Are we at the point yet were simple HA meaures are basically free? I mean, I don't manage the servers where I work, but even I see that virtualization make a lot of maintenance shit really trivial, so you skip the step of "upgrading from a hard server to VMs" immediately and having "no HA measures in place" seems ludicrously incompetent.
I'm not sure if I made myself clear.I'm not sure I understand. Some HA stuff is cheap--it's pretty easy to set up parallel web/application/DNS servers behind a load balancer. It's pretty easy to have hot spare load balancers.
Machines that hold state are harder. There are all sorts of replication/clustering strategies, all of which have benefits and disadvantages. But, basically, you can set up something like MySQL to do semi-synchronous, multi-master replication and if the primary fails you should be able to failover to the secondary quickly. Depending on how your software works, it may throw up an error because of the failover or it may just pause for a few seconds until it can reconnect to the secondary.
Replicating an NFS server with FOSS solutions is a real PITA. Proprietary NFS devices can usually do it well. NFS is kind of shitty, though, so I'd carefully weigh my use case. If I'm dealing with a bunch of tiny, few kb files, I might just stash them in a database field. (And before you say "WTF!!" remember that databases handle failover and replication really well.) If you have large files (which is probably the case because we're probably talking about user-uploaded media files) I'd use a distributed file daemon that writes the files to several nodes; there are lots of them. You lose POSIX semantics, but you probably didn't need those anyway.
Network connections can be HA'd pretty easily--most servers have multiple ethernet ports and it's pretty easy to set up multiple connections to multiple switches and then use either failover if the primary switch dies, or bonding to just use both switches simultaneously (which gives you a nice throughput boost, too, since you can saturate several connections).
Anyway, yeah, HA isn't that hard, it just requires forethought.
-
@Cassidy said:
@morbiuswilters said:
Eh, half the systems I designed you'd reboot for a config change or software change.
Do you actually mean reboot, or do you mean restart/reload that service? I'm used to the latter in the *nix world.
Reboot. The machines boot off the network using a standard image. It's simpler to just build a new image, then rotate through servers rebooting them so they pick up the new image. That way I don't have to come up with a way to apply changes to already-running machines. Also, I guarantee that any given machine has been booted using the exact current image it has; no problems with updating 100 machines while they are running, then having a power outage and discovering they can't boot using the image. Like I said, reboots are no big deal. Rebooting once a day is nothing in a well-designed, highly-available farm.
-
@PJH said:
@da Doctah said:
or even things that don't have a K in them at all, like potassium
Kalium. Like Tungsten (wolfram), Lead (plumbum), Tin (stannum) - there are reasons for them. Much like the names given to Linux packages/software. Though the latter tend to be less obvious (or non-existant)... @da Doctah said:thousand
Kilo...@da Doctah said:absolute temperature
Kelvin.
I don't have previous knowledge of the others you listed, so will plead ignorance on them.K for strikeout comes from baseball scorekeeping, sometimes printed mirror-imaged. The folks at Circle K did a tie-in using their logo a few years ago.
K for black is from four-color or CMYK printing; it actually stands for "key", but there are already enough bogus color names in the LL Bean catalog (what the hell is a "loden"?).
Phylloquinone is the chemical name for vitamin K.
K for lysine comes from protein sequencing; the letter L was already taken for leucine. (W is used to stand for tryptophan.)
And K for kilo (original Greek χίλια, so it should really be "Ch") means both 1000 and 1024, sometimes in the same sentence.
-
@spamcourt said:
@morbiuswilters said:
@dtech said: Kernel updates are updates, dumbass. [...] Did you post just to display your ignorance or did you have a point?
Could you please not insult people all the time? Yes I know you disagree with what he said, and yes you do have a point, but that's quite annoying and unnecessary, and it actually makes these forums a worse place for everyone.
I'm sorry it bothers you (really, I am) but I think it makes things more unpleasant for people to shout "HA HA, got you!" when they think they're correcting me, when in truth they are wrong. I mean, I said that a system will need to reboot because of updates and he comes back with "omg u r so rong, thats only kernel updates" which manages to miss the entire point and be pedanticly dickweedish while also being incorrect.
-
@da Doctah said:
And K for kilo (original Greek χίλια, so it should really be "Ch") means both 1000 and 1024, sometimes in the same sentence.
No, it doesn't. Kilo always means 1000; 1024 is always kibi. People may use it incorrectly, but they are wrong.
-
@morbiuswilters said:
@da Doctah said:
And K for kilo (original Greek χίλια, so it should really be "Ch") means both 1000 and 1024, sometimes in the same sentence.
No, it doesn't. Kilo always means 1000; 1024 is always kibi. People may use it incorrectly, but they are wrong.
Just because it's correct doesn't make it right.
-
@boomzilla said:
Just because it's correct doesn't make it right.
Just because it's in poor taste doesn't make it wrong.
-
-
@mott555 said:
So a bit like a real hammer then?If Microsoft manufactured carpentry tools and Visual Studio was a hammer, it would
- Have several models depending on how big of a nail you want to use
-
@Cassidy said:
@morbiuswilters said:
Eh, half the systems I designed you'd reboot for a config change or software change.
Do you actually mean reboot, or do you mean restart/reload that service? I'm used to the latter in the *nix world.
Indeed. After an update you indeed need to restart the deamon (service), but you don't need a full reboot. That's why it's not uncommon for unix systems to have uptimes of several months.
-
@dhromed said:
http://www.youtube.com/watch?v=fX56D9GzCQA@Xyro said:
Am I funny yet? Maybe if more people contributed to this pattern, it would retroactively become funny.
I wish I could permaban your dad's dick from your mom's vagina, before you were conceived.
-
@morbiuswilters said:
which sounds hacky enough that I'm pretty sure more organizations wouldn't bother with. Swapping out the kernel while it's running sounds like a good way to inadvertently corrupt data.
OK, so you're basically biased against it because it sounds hacky? Have you even taken a proper look at it? You are offering plenty of opinion and little in the way of experience and fact, par for the course for you it seems :)
-
@morbiuswilters said:
I'm sorry it bothers you (really, I am) but I think it makes things more unpleasant for people to shout "HA HA, got you!" when they think they're correcting me
Actually, I doubt you're sorry, else you wouldn't continually be so downright rude to people all the time. And it's only unpleasant for you to be called out when someone thinks you're wrong, but it's often unpleasant for more than the recipient of your abuse when you hurl it. Some animals throw shit at each other in an attempt to prove superiority, I suggest you evolve.
-
@ASheridan said:
And it's only unpleasant for you to be called out when someone thinks you're wrong, but it's often unpleasant for more than the recipient of your abuse when you hurl it. Some animals throw shit at each other in an attempt to prove superiority, I suggest you evolve.
TDWTF just wouldn't be the same.
