Compare ClearOS with Zentyal
-
@scottalanmiller said in Compare ClearOS with Zentyal:
So let's just look at Windows GUI or Webmin or Zentyal... I've not tested this so I'm really asking. If you run out of inodes on any of those systems.... how does the GUI tell you that inodes are exhausted? Since this is a GUI vs CLI question. I have no idea how a GUI would not make things worse here, I've seen no GUI report on inodes, but the CLI does.
I've never had the error until now. I believe I was just running an
apt-get upgrade
and it failed simply saying I was out of space. So a quick check withdf
showed I wasn't. Of course I learned about the inode switch later, but my only point is that there was nothing intuitive about the situation. I had to Google why it was saying out of space when I wasn't. It didn't take long to find the inode angle.I don't know how the GUI would handle it. Since most GUIs would have some interface for doing software updates, I would imagine I would also just get a visual error about being out of space.
The question, is the GUI/software smart enough to take it a step further and tell me why I'm out of space, or just leave it at the error?
My guess is the GUI would not be smart enough to take it further, and simply error about being out of space and leave it at that.It's no different in Windows when you try to install something and you just get some random obscure error and it closes.
I remember back in my computer repair days, software would sometimes throw DLL errors and we'd have to reregister the libraries using CLI commands.A GUI doesn't help with some stuff, but it does with others. Using GUIs for configuration, where you just have to look at checkboxes, drop down lists, toggle switches, text fields. So much easier than running around the file system editing touchy text files. Editing my network in Windows is worlds more intuitive than messing around with interfaces and network files.
Finding large files in Windows is just a matter of clicking the "size" filter in the search box and choosing "gigantic". It won't limit to 10 results though, it will just show everything. Or I use a nifty utility called WinDirStat.
-
@guyinpv said in Compare ClearOS with Zentyal:
The question, is the GUI/software smart enough to take it a step further and tell me why I'm out of space, or just leave it at the error?
My guess is the GUI would not be smart enough to take it further, and simply error about being out of space and leave it at that.That's my thought and that would make it SO much worse when the GUI just lacks the info to explain what has happened.
-
@guyinpv said in Compare ClearOS with Zentyal:
Finding large files in Windows is just a matter of clicking the "size" filter in the search box and choosing "gigantic". It won't limit to 10 results though, it will just show everything. Or I use a nifty utility called WinDirStat.
Oh okay, search does that. Makes sense. I would not have looked there.
-
@scottalanmiller said in Compare ClearOS with Zentyal:
@guyinpv said in Compare ClearOS with Zentyal:
Anyway that's neither here nor there. I just think some people like the CLI environment, some people get it, some people "click" with it. But my brain just has a hard time. If my work is 80% Googling and 20% CLI work, but using a GUI makes it 80% work to 20% researching, I'd rather take the GUI. I just get more work done.
I would agree with that. The problem is, once you get to any scale, the GUI is a huge problem. I've never seen anyone break the 60 servers per admin mark using GUIs. Even six figure people. They always get mired in the interfaces and the slowness of it all.
On the CLI, you can scale up like crazy. Even without special tools. Add in the special tools and you can be at thousands of machines per person.
I'd like to see this in operation. I mean, what are they actually DOING with the servers in the first place? Why would they manage all 60 at once? Wouldn't you have some sort of management utility on them which report to a central interface? Then from the central interface you can hop in and do work on any given system?
What would be an example where managing 60 servers fails but using CLI tools works well? Just some random scenario, so I can wrap my head around it.
-
@guyinpv said in Compare ClearOS with Zentyal:
I'd like to see this in operation. I mean, what are they actually DOING with the servers in the first place? Why would they manage all 60 at once? Wouldn't you have some sort of management utility on them which report to a central interface? Then from the central interface you can hop in and do work on any given system?
So I've been this guy, both in the 600 snowflakes world and in the thousands of DevOps style world. You really do do this stuff.
It's not like "hey, I'm going to deploy this to 600 boxes right now". But I might deploy software to 100 boxes at a shot on a Friday night. In the snowflake world, no management utility at all. Just SSH and a Bash shell. It's so easy nothing more is needed. You can loop or use DSH style stuff.
In the DevOps world, there is a "central console" but it is still command line. You use tools like GIT to hold configuration information, you might use a minimal GUI for groups and stuff, but only that.
-
@guyinpv said in Compare ClearOS with Zentyal:
What would be an example where managing 60 servers fails but using CLI tools works well? Just some random scenario, so I can wrap my head around it.
You mean where 60 fails from a GUI?
Let's say you need to deploy a package to 60 servers during a 20 minute maintenance window. With windows you could do this with external third party tooling or PowerShell pretty easily. But from the Windows GUI? How do you do it?
-
@scottalanmiller said in Compare ClearOS with Zentyal:
@guyinpv said in Compare ClearOS with Zentyal:
What would be an example where managing 60 servers fails but using CLI tools works well? Just some random scenario, so I can wrap my head around it.
You mean where 60 fails from a GUI?
Let's say you need to deploy a package to 60 servers during a 20 minute maintenance window. With windows you could do this with external third party tooling or PowerShell pretty easily. But from the Windows GUI? How do you do it?
You got me, I don't know. I was asking YOU the question. How do admins typically try and manage 60 servers with a GUI? Aren't there any standard tools in the corporate world for this?
In my mind, I wouldn't be just using the Windows GUI. Theoretically I'd have some kind of management interface which is connected to all my systems. As long as I had the authority to install packages, I can see myself just selected which systems to install to, picking the package from some central data store, then clicking the "DO IT" button and sit back and relax while it all happens.
Either that sort of interface simply doesn't exist, or I don't know how else to do it except by connecting to each system one by one.
See in my mind I think a GUI can do ANYTHING a CLI can, since a GUI is simply abstracting the same commands into a visual interface. All the GUIs do at the end of the day is run commands to the underlying APIs, but provide an interface to abstract the ugly stuff away.
When I open Windows Explorer and browse to a folder, under the surface it's just running the same sorts of file system commands I can type myself, it just shows me the results more pretty.
I would think the same is true for managing more than 60 servers. Whatever you can type in a CLI to manage them, the GUI would be doing the same, but showing me the results visually and abstracting some commands into buttons and checkboxes.Hey, I don't know. If the tools I'm imaging don't exist yet, then there is a market for it!
-
@guyinpv said in Compare ClearOS with Zentyal:
You got me, I don't know. I was asking YOU the question. How do admins typically try and manage 60 servers with a GUI? Aren't there any standard tools in the corporate world for this?
Oh. then the answer is no. I've never worked in any environment of any size that used anything but the command line. GUI for a one off server, I totally understand the logic and the reasoning. But at scale, GUIs are basically out of the question. Always have been. Never seen anyone even think to ask for one.
I've worked in environments of all sizes and all kinds of industries. One off Windows machines used to be all GUIs, even that is slowly changing. But UNIX systems of any sort, 100% CLI. I've heard of people with just one or two servers trying to use GUIs. But I've never heard of a shop with more than a few and have never encountered first had a shop with even only one, that used a GUI.
I can't imagine how using a GUI would work, realistically. I mean we use them for log reading (ELK) and monitoring (Zabbix) but those are decoupled systems, not the boxes themselves.
-
@guyinpv said in Compare ClearOS with Zentyal:
In my mind, I wouldn't be just using the Windows GUI. Theoretically I'd have some kind of management interface which is connected to all my systems. As long as I had the authority to install packages, I can see myself just selected which systems to install to, picking the package from some central data store, then clicking the "DO IT" button and sit back and relax while it all happens.
These exist for package deployments, especially for desktops. but once you are managing the boxes, you pretty much need fast logins. Even Windows I find remoting into a GUI to be painfully slow. i can often do a task in Linux before you can get the RDP connection made for Windows.
And I'm only half kidding.
-
@guyinpv said in Compare ClearOS with Zentyal:
Finding large files in Windows is just a matter of clicking the "size" filter in the search box and choosing "gigantic". It won't limit to 10 results though, it will just show everything. Or I use a nifty utility called WinDirStat.
That can be done with
ls -lSh
. Lists all the files in that directory, sorted descending by human readable size.How do you remove all of the files in a directory with a certain extension in a GUI?
With a Linux CLI it's
find . -type f -exec rm -f {} \;
I can type that 100 times faster than I can ctrl select each file in a GUI and hit delete. -
@guyinpv said in Compare ClearOS with Zentyal:
See in my mind I think a GUI can do ANYTHING a CLI can, since a GUI is simply abstracting the same commands into a visual interface. All the GUIs do at the end of the day is run commands to the underlying APIs, but provide an interface to abstract the ugly stuff away.
This is absolutely true. What they can't do is interact with the humans as quickly and easily. Microsoft tried this for a generation and finally gave up and made PowerShell the master administration tool for Windows. The UNIX world just crushed them in productivity no matter what they tried on the GUI side, and MS has some great GUIs.
In the end, humans just proved to be way more efficient with CLIs when working on systems regularly.
-
@scottalanmiller said in Compare ClearOS with Zentyal:
@guyinpv said in Compare ClearOS with Zentyal:
What would be an example where managing 60 servers fails but using CLI tools works well? Just some random scenario, so I can wrap my head around it.
You mean where 60 fails from a GUI?
Let's say you need to deploy a package to 60 servers during a 20 minute maintenance window. With windows you could do this with external third party tooling or PowerShell pretty easily. But from the Windows GUI? How do you do it?
The only thing I can think of is something like Dell Kase where you can tell it to deploy software on certain systems, but it's still slower than CLI.
-
Ah didn't make it down far enough to see you discussed those already.
-
@johnhooks said in Compare ClearOS with Zentyal:
@scottalanmiller said in Compare ClearOS with Zentyal:
@guyinpv said in Compare ClearOS with Zentyal:
What would be an example where managing 60 servers fails but using CLI tools works well? Just some random scenario, so I can wrap my head around it.
You mean where 60 fails from a GUI?
Let's say you need to deploy a package to 60 servers during a 20 minute maintenance window. With windows you could do this with external third party tooling or PowerShell pretty easily. But from the Windows GUI? How do you do it?
The only thing I can think of is something like Dell Kase where you can tell it to deploy software on certain systems, but it's still slower than CLI.
There are some tools out there for the GUI stuff, I just have never seen them and am not aware of any that handle ad hoc tasks, which is really where the CLI shines.
-
Let me give an extreme management example. Of course, this like never comes up, but it's an example.
Today @guyinpv discovered that he is out of INODES on one box. Crap, he thinks, what if I'm almost out of INODES on every box!
So he goes to check. He pours through GUIs trying to find this info. Likely, no GUI anywhere for any system shows it. Certainly no normal one.
What do we do on the CLI? The same thing that we do for normal file system checks except with the addition of the "-i" flag.
for i in $(cat serverlsit); do ssh $i "df -i" ; done
Boom, we have a list and we can see if anything is getting dangerously high.
-
@scottalanmiller said in Compare ClearOS with Zentyal:
Let me give an extreme management example. Of course, this like never comes up, but it's an example.
Today @guyinpv discovered that he is out of INODES on one box. Crap, he thinks, what if I'm almost out of INODES on every box!
So he goes to check. He pours through GUIs trying to find this info. Likely, no GUI anywhere for any system shows it. Certainly no normal one.
What do we do on the CLI? The same thing that we do for normal file system checks except with the addition of the "-i" flag.
for i in $(cat serverlsit); do ssh $i "df -i" ; done
Boom, we have a list and we can see if anything is getting dangerously high.
In my theoretical world, I would never have to "wonder" if inodes are running out. There would be some sort of server management tool on all my servers, which report back to me any kind of system level problems. Something like Spiceworks for example. Reports when drive space is low, printer is out of ink, CPU usage is high, RAM is running out, too many connections, failed log in attempts, log errors, root level log in attempts, etc etc.
So I'm sitting at my workstation and I want to check inode and drive space, I would simply open "that part" of my GUI management interface, go to drive space, and sort by inode or free space or whatever to see where I'm at.
The command you typed is little more than "show me inode usage on all my systems", which could easily just be a function somewhere in the GUI of some server management interface.
I can't believe stuff like this doesn't exist!
In your scenario, if you had not had the intuition to type that command and look, how would you know any server was running low in the first place? Wouldn't you want something like that to be monitored and reported back to a central interface?
-
@guyinpv said in Compare ClearOS with Zentyal:
In my theoretical world, I would never have to "wonder" if inodes are running out. There would be some sort of server management tool on all my servers, which report back to me any kind of system level problems. Something like Spiceworks for example. Reports when drive space is low, printer is out of ink, CPU usage is high, RAM is running out, too many connections, failed log in attempts, log errors, root level log in attempts, etc etc.
I appreciate the value of the theoretical world and all of the information that could potentially be collected. Such tools can exist, but are often extensive and intrusive if used all the time, there is just so much data. If you turn all ad hoc monitoring into full time monitoring you could, in theory, do this. But it is hard.
There are no products that do this or try to do this, probably for a reason. If you did, the GUI would be so impossible to deal with that no one could handle it. There is just so much potential data.
-
@scottalanmiller
There is a lot of "potential" data, and a ton of arbitrary data. But only so many critical error conditions relating to that data.
Testing drive space would be one of those daily or weekly checks. Testing inode space could be monthly or whatever. It's not like you need a constant stream of reporting. Only if it's critical levels would anything get reported.What is the alternative then? To avoid monitoring too many variables, we just sit around and wait for problems then scramble trying to figure out what happened?
Don't you even monitor basic stuff like CPU/RAM/HDD? Don't you get alerts on successful root logins? Do you monitor anything at all?
Just to be random, would you stick even something like Newrelic on servers? How do you even know if a server goes down without some kind of monitoring/management tools? I assume maybe you use Nagios or something. Those can certainly record a ton of data!
In any case, the GUIs I'm talking about are not on the servers themselves. I'm talking only about having a GUI to manage them. Most likely from your own workstation or even a web interface.
In my case, ClearOS is just like Windows, where the GUI and services are pretty much intertwined. For Zentyal, it's installed as software on top of the server. But for managing 60 servers, ya there is no way I'd want 60 logins to Zentyal! In that scenario, I would only have some type of management/control software on each server which lets me do everything remotely from some other central interface. Not unlike what XO does, but those types of tools don't have features to let me run things across multiple servers.
I really am surprised there are not good options here.
-
@guyinpv said in Compare ClearOS with Zentyal:
@scottalanmiller
There is a lot of "potential" data, and a ton of arbitrary data. But only so many critical error conditions relating to that data.
Testing drive space would be one of those daily or weekly checks. Testing inode space could be monthly or whatever. It's not like you need a constant stream of reporting. Only if it's critical levels would anything get reported.You can do that easily today with a CLI tool
For you it might be a monthly thing, but for a lot of companies this is a daily thing. A one size fits all GUI would be tough.
-
@guyinpv said in Compare ClearOS with Zentyal:
I really am surprised there are not good options here.
Well on one side you are recognizing an opportunity. That's good. But from the system admin side, I don't see a problem to solve. Already everyone in an environment of any scale has a system that works incredibly well, the command line. There isn't a problem to solve. So while you might have a solution, it's a solution in search of a problem.
not that the idea is bad, but attempt to implement it, even on paper. How man of these ad hoc things do you want to include? Any one you leave out might be your "inode" to someone else. What does this GUI based system really look like? how do you make an interface for "everything"? And then make it easy to use.