Best posts made by matteo nunziati

gluster has been choosen upstream and by read hat for their hyperconverged stuff, so it should be ok, even if use it without ovirt seems a bit complex. But it is just a feeling, I never used it, just had a look to the how-tos

@scottalanmiller said in Is this server strategy reckless and/or insane?:

@matteo-nunziati said in Is this server strategy reckless and/or insane?:

@dustinb3403 I'm running it on microSD .... Brrrr

So many posts... running what?

Running hyperv on microSD. HPE microSD.

@lakshmana cwntos has a wiki page on the issue

@networknerd currently all sql dbs do in memory cache AFAIK therefore if your dataset is not really sparse you should be served well

@creayt said in Is sharing a single network connection between two servers dumb?:

I guess what I'm really looking for I guess is how much of a performance hit will the 2nd ( daisy-chained ) server have for internet connectivity, now that I know it's possible technically?

Mostly will depend on how much of the connection workload will be put on main sys: if you are querying and the cpu if 100% and network load will hit it, well you will have impact. Remaining stuff should be neglectable.
Can't think of anything at least.

actually you should keep your backups both in and out of your facilities.

then, what kind of topology do you have? are you backing up clients or is this a database?!
which price range? you are talking about NASes. If you mean those "cheap" appliances I think you can think about a cloud solution like ...

hey @coliver just give me the time to answer!

+1 for backblaze B2, just thinking about same thing. It has retention therefore you can backup stuff and keep copies

OR

if your data are highly compressible just compress them and backup with stuff like veeam agent on a local proxy and then push on B2 without retention (veeam will keep retention)

to be clear:
in python 2.x what you get with a print() is a tuple of elements. the comma is there NOT because it is a separator but because it is used in print() to concatenate multiple items (Python objects) in a tuple.

in python 3.x the print() statement print EVERYTHING if it is a string and concatenate the objects into a single string. If you do not pass a string but an object has a __str__() method, it is implicitly invoked.

trivial example in python 2, create a tuple:

t=("rob",21)
print(t)

try to print it. you get exactly the same result than your example. then try the same in python 3.x. again...

the comma in the print statement is misleading you. it is just command synthax nothing to do with a char in a string or anything similar. when you see stuff into rounded brackets, this is a tuple. and by default elements in a tuple are separated by a comma.

(to be accurate the representation of a tuple according to its default __str__() method)

NOW
try the same thing calling tuple items explicitly:

print(t[0], t[1])

output is different in python3! In this sense pythion 3 is more accurate in the data representation.

EVENTUALLY
if you really want a comma separated list of items into a string you must format it:

 print("%s; %s" % t)

in this case I've used a semicolon to point out the difference.

@black3dynamite said in KVM Virtual Machine Boot Order:

Uncomment START_DELAY=0 in /etc/sysconfig/libvirt-guests file.

https://access.redhat.com/discussions/685243

This will kill parallel startup adding a delay between each vm boot

@scottalanmiller said in Best practice on IPMI/iDRAC/ILO?:

@pete-s said in Best practice on IPMI/iDRAC/ILO?:

How about the machines that doesn't have a dedicated port?

This CAN be VLAN'd but makes doing so much more of a pain and defeats much of the purpose. This is one of the many reasons that people rarely do this today.

Usually shared ports can be tagged from within the OOB interface. ILO definitively lets you apply a vlan tag.

any L2 managed switch with spanning tree?

Spped is mostly about local cache. Even docker requires you to bootstrap the host and usually you do it with ansible like tools. Then you download the docker images and so...

What is attractive about stuff like docker is the automation-by-default approach which has created a really good ecosystem with a lot of premade automation already done.

Doker images on kubernates are awesome but if you have to manage your data center doing ansible + docker over just ansible is not gaining a lot imho.

If docker works out of the box it saves the initial time required to write ansible scripts and design them

backup days

CONTEXT
It is almost a year since I changed job and now I'm a freelancer with my trusted Thinkpad T440 (lenovo and trusted in the same sentence XD).

As a freelancer I'm asked to develop a lot of code for .NET so I've switched my main OS to windows (due to Visual Studio Code 2017).

BACKUP NEEDS AND BACKUP CHAIN
I've installed the free version of Veeam Agent and I'm backing up my entire computer. Let me explain the goal of the backup first:
1- all of my source is hosted online in bitbucket/git
2- all the docs are online in gdrive
3- basically backups are required to speedup windows restoration in case of disaster as it always requires tons of hours to setup a windows OS with all the required applications (linux is way faster in this)
4- backup jobs run once per week

My current backup chain is: laptop w/ veeam => home NAS (small 2bay , slow cpu) => backblaze B2 sync (integrated into the NAS)

THE ISSUE
The chain is ok but the veeam settings are killing my internet connection without providing real time advantanges: I've a 3Mbps upload speed.

Veeam is currently set up to take a full snapshot of the laptop and do incremental backups (with dedup and compression set up for WAN). Retention is 14 backups. The issue is that when you take the 15th+ backup, you start merging the base snapshot forward.
1- Merging the snap is as long as do a full bakcup (usually around 1:30 hours or 1:45)
2- once the baseline snap is updated I need to reupload to B2, which means that my NAS is still moving something like a whole new snap of the laptop (baseline snap + new increment). which means soething around 100GB.

MY IDEA
As backups aim is mostly for fast windows restoration I was thinking about a couple of things:
1- as incremental backups still require a lot of computation time and a lot of upload time, maybe it would make more sense to simply do full backups every time.
2- as I'm just snapshotting the laptop to speed up restoration I could reduce the frequency to 1 backup every 2 weeks giving a bit of "rest" to my uploads.

YOUR OPINION
Can you suggest anything else?!

@dbeato said in Rethinking my backup strategy:

@JaredBusch said in Rethinking my backup strategy:

@scottalanmiller said in Rethinking my backup strategy:

@JaredBusch said in Rethinking my backup strategy:

@scottalanmiller said in Rethinking my backup strategy:

@matteo-nunziati said in Rethinking my backup strategy:

@scottalanmiller said in Rethinking my backup strategy:

@matteo-nunziati said in Rethinking my backup strategy:

@Obsolesce @black3dynamite so this chocolatey could be the missing bit to speedup redeployment of a win machine as I do with linux+ansible+current distro's repo... I will check it.

Oh it's amazing.

If you have Ansible already, I'm with the "only backup once in a while."

Or ever, is there really a need to backup at all?

Well Windows patching and applications install is a pita. I will look into chocolatey. Wasting a day in reinstalling stuff in attended mode is not on my plans...

Do you have a lot of apps to install? VS is a big one, that should be able to be automated.

Would storing a copy of installers help?

VisualStudio Community is in Chocolately actually.

@JaredBusch said in Rethinking my backup strategy:

@scottalanmiller said in Rethinking my backup strategy:

@matteo-nunziati said in Rethinking my backup strategy:

@scottalanmiller said in Rethinking my backup strategy:

@matteo-nunziati said in Rethinking my backup strategy:

@Obsolesce @black3dynamite so this chocolatey could be the missing bit to speedup redeployment of a win machine as I do with linux+ansible+current distro's repo... I will check it.

Oh it's amazing.

If you have Ansible already, I'm with the "only backup once in a while."

Or ever, is there really a need to backup at all?

Well Windows patching and applications install is a pita. I will look into chocolatey. Wasting a day in reinstalling stuff in attended mode is not on my plans...

Do you have a lot of apps to install? VS is a big one, that should be able to be automated.

Would storing a copy of installers help?

VisualStudio Community is in Chocolately actually.

Oh wow, cool. Had no idea.

So is SQL Server Express.

That is how I seutp my Win 10 VM for some dev stuff.

and many other things! IT has been awesome ever since I found Chocolatey!

So I've tested choco in a vm. So far so good. I need to be acostumized to the cmd line and terminology. As a side this has pointed out how slow and huge win packages are... Used to apt-get this is super sloooooow.
They also apply kb fixes in the process and this is c00l!

I'll dig a bit in their repo... but this could resolve part of my problems.

@scottalanmiller with any devel tool and namely with more "modern" WPF it is trivial to make big controls which scale with window size or screen resolution.
Simply a lot of SW is badly designed.

@EddieJennings said in KVM / Red Hat Virtualization Management:

@black3dynamite said in KVM / Red Hat Virtualization Management:

oVirt web management UI looks very similar to what RHEV uses.

I'm surprised Red Hat doesn't seem to be promoting whatever they use for managing RHEV. I figured they'd try to lure folks away from VMware or Hyper-V with a slick interface

Ovirt is the upstream of RHEV.

Always default but I move asap to dark themes.

what a mess...

I'm not a forensic expert in digital forensics, anyway all starts with a proper chain of custody and an unaltered probing env, which usually means you have to perform low level dumps of stuff you intend to manage, in order to avoid to leave your traces over the previous traces (in classic HDD you want to ddrescue everything... if you can go offline).

so NO, you cannot use any account feature to demonstrate anything, unless you want your activity to be discarded (at best)

nowdays I can only listen to music when I'm alone in my car: I've an mp3 collection on an usb pen. Not allowed to listen to music in office, can't listen to it when I'm with my wife and daughter... I'm unable to listen to music and do sport: I'll end following music rhythm and not the required one.
Basically I've music only on commutations.

@stacksofplates maybe of interest: my libvirt backup wrapper