warren's questions

When deciding whether to use LVM volume groups or ZFS pools in configuring new file servers, what needs to be considered? Is there a "better" choice in the realm of multi-purpose file servers? Unlike this previous question, I don't want to layer the technologies.

Scenario:

• RHEL / CentOS 6 x64 servers
• many available, identical DAS and SAN LUNs

Choice:

I am personally quite familiar with LVM, so am comfortable using it if it is the better option. However, ZFS looks pretty promising, and learning new technology is always good.

Given that we want to be able to share-out a fairly large store (multiple TB) to different departments, and they need to be able to access them over both CIFS and NFS, should we use ZFS or LVM on for the underlying pool manager?

^{_{I know that using a product like FreeNAS is possible, but for a variety of reasons, I need to be able to roll-out onto "typical" Linux servers.}}

When building a VM, you can select which virtual device type you would like a vNIC to be (E1000, VMXNET3, etc).

After the VM is created, can you change the type of vNIC in use on a given connection (eg from E1000 to VMXNET3)?

If so, how?

How can I suppress giving a reason for shutdown on a Windows Server host?

Specifically, on 2008 R2, but all versions back to 2003 and up to 2012 would be appreciated.

When a process spawns a subshell, how is the ulimit inherited?

Is it by user, by shell, or a combination?

Specifically, this was found in the context of HP Server Automation jobs failing on some managed AIX devices because of an interaction between the soft and hard ulimits in place. In short, the Agent (which runs as root and generates a subshell when processing commands from SA) was unable to process some jobs because its subshell was hitting a ulimit.

Increasing the ulimit solved the problem, but the question of how the ulimit is inherited, and whether the soft or hard (or both) limits need to be raised is left open.

What is the appropriate answer to this question? It may be AIX-specific, or it may be across Unix/Linux variants - whichever is better.

Based on a previous question, I installed ipmitool (yum install ipmitool).

Even after a reboot, though, i get the following error when trying to run ipmitool power status:

Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory
Unable to get Chassis Power Status

Is this an OS/hardware issue (CentOS 6.3 x64 on a hosted machine in a remote datacenter - unsure on hardware vendor)? Or have I missed something more elemental in installing ipmitool?

Slightly related to a previous question about thick vs thin provisioning, why does eager zeroing the to-be-cloned VM take longer than lazy zeroing?

For example, in a recent test we ran, an eager-zeroed clone took ~85 minutes to complete, whereas changing to lazy zeroing made it complete in ~50 minutes.

How much overhead is incurred when eager zeroing vs lazy?

When would 'eager' be the right mode to use?

Several applications require a specific release of Linux on which to run, such as RHEL 5 x64, Update 4.

Theoretically, CentOS 5.4 x64 should be exactly the same as the RHEL variant - since the CentOS project endeavours to be 100% binary compatible with RHEL.

Likewise, Oracle Enterprise Linux should be identical - but many applications will fail to work properly when not running on the "proper" distribution.

The cheap fix of just installing the appropriate redhat-release rpm is not always effective either.

Under the hood, what are the real differences between RHEL and the repackaged editions?

I typically use thin-provisioning of storage space when building VMs: it provides a lot of flexibility, and seems to be faster to build.

The only potential definite advantage I can see is to ensure that a VM doesn't accidentally run out of available space on the underlying storage infrastructure - are there others?

When sending large email to a new CentOS6 server running Postfix as the MTA, the following message is returned:

tried to deliver your message, but it was rejected by the recipient domain. We recommend contacting the other email provider for further information about the cause of this error. The error that the other server returned was: 552 552 5.3.4 Error: message file too big (state 18)

I found the following suggestion, but am unclear as to where it needs to be added in the main.cf file:

This was caused by Postfix and it's limit on not only messages but mailbox sizes.

I had to add this setting in /etc/postfix/main.cf:

message_size_limit = 31457280

How can the maximum mail size (including attachments) be increased in Postfix?

The last real question I've seen on this topic is from about two years ago (is ext4 ready for production usage).

In the intervening time, how has ext4 improved?

XFS, JFS, and ext3 are the standby reliable choices. I've only used ext4 on recent Ubuntu test/dev environments, and haven't seen any issues - but they're also low-use workstations, VMs, and throw-away training environments.

How does ext4 stack up now that it's had [some] time to mature vs XFS and JFS (especially) from a speed and reliability standpoint?

Is ZFS a viable option (seeing as it's a fuse module, probably not for Linux - yet)?

I have noticed recently that sometimes tail -f <logfile> will stop updating to the screen.

Doing a Ctrl>-C and restarting the tail works fine, though. And I checked to make sure the logfile isn't being rotated midstream (which can make tail lose its mind).

What would cause this? I'm running RHEL 5.2 x64.

I our lab environment I have several Win2k3 and a couple Win2k8 servers.

I don't really care why the servers reboot, and would like to be able to disable the shutdown dialog (and the "why did this server shutdown" on reboot in the event of power failure).

Is this possible, and, if so, how?

How can passwordless sudo access be setup on either RHEL (Fedora, CentOS, etc) or Ubuntu distributions? (If it's the same across distros, that's even better!)

Setting: personal and/or lab/training equipment with no concern for unauthorized access (ie, the devices are on non-public networks, and any/all users are fully trusted, and the contents of the devices are "plain-vanilla").

When do entries in cron.daily (and .weekly and .hourly) run, and is it configurable?

I haven't found a definitive answer to this, and am hoping there is one.

I'm running RHEL5 and CentOS 4, but for other distros/platforms would be great, too.

From an end-user perspective, what is the difference between a NAS device and using NFS exports from a file server?

They seem to accomplish the same end result.

The difference between a SAN and other file storage is related (in my experience) to how they are connected to the server infrastructure.

However, the difference between a NAS, connecting over a standard ethernet port, and NFS (sharing storage off specific servers, also over the network), seems more nebulous.

Is there a good reason to pick a NAS filer over just running NFS on servers?

I'm sure I've just missed which tutorial/manual page covers this, but how do you add get the guest OS to recognize that you've added new drives to it without a reboot?

I have a RHEL5 guest running on ESX 4. I've added new virtual disks to the VM, but have not figured-out how to get the guest to recognize them without a reboot.

Is this possible? If so, how?

Specifically, I'd like a way to do this in RHEL, but any other Linux distribution or edition of Windows would be nice, too.

If I have multiple hosts configured on one machine (a la apache's VirtualHosts), how can I do a lookup on the IP and find all domains configured to reach it?

For example, I have several web and email domains hooked-to my server. How can I find all domains that point to it?

Is it even possible?

I have DNS A entries for all the domains I own, plus I know some friends' domains point to my server. What I'd like to see is if folks I don't know about are pointing there, too. (Or if someone has repointed their domain elsewhere, and I can delete their 'old' website from my server.)

Based on this question regarding enhanced permission support in filesystems, what is the best permissions + directory and mail stack to compare against Active Directory on NTFS with Exchange?

I'm familiar with OpenLDAP for directory services, and Scalix for email and calendaring. Are there other good competitors in this arena? Novell's NetWare would have been my answer a decade ago, but with that having pretty much disappeared, I haven't seen a good replacement, outside of "perl-and-duct-tape".

ie - is there an integrated packaging of these (or other) toolsets?

The thrust of this question is to see how plausible a completely AD-free environment is, with a quality calendar, email, and directory services functionality.

It can run on Mac OS X, Linux, or Windows - I'm just very interested to see what others have done/seen/used in this area.

In a similar vein to this question, how would I do a schema-only dump in PostgreSQL?

I've been working with HP's Server Automation tool suite for a long time, and while it's a great tool - it's too much for a small business or home network to use (more accurately: it's way too expensive for anything less than a large company to afford).

I'm somewhat familiar with Puppet and OpenNMS for provisioning and monitoring. After googling for other information, though, it seems there's not much in the way of management tools that are mature enough to use without a lot of effort to maintain them.

Are there any other FLOSS tools (commercially supported is fine) for managing:

• server provisioning
• patch control
• software package pushing
• etc

Expansion

• integration with native provisioning is fine (kickstart, jumpstart, autoyast, unattend.txt, winpe, etc)
• a GUI of some kind would be nice, but not required
• I want to be able to setup, for example, an OS profile, and ship it to X-many devices, preferably with fields that would be filled-in during the process for hostname, IP, etc
• agent-based is preferable to non-agent-based (ie, I know I could use soemthing like pssh for package and patch management, but that requires pre-shared keys as opposed to agent/certificate authentication)
• being able to tie-in to Nagios, OpenNMS, or similar would be fantastic - so known ports, processes, and SNMP data would be great
• virtualization support would be really helpful - but I'm OK with it not being able to create the VM, just populate/provision it

If the answer to this question turns out to be, "here's a bunch of tools that are part of the solution", I'm willing to build a wrapper of some kind for my own use.