For ArchLinux, I've seen this package surface:
openssh-xattr
As it seems a patched openssh that adds xattr facilities to the sftp ssh internal server. What's going on with this patch? Will it arrive on Linux any time soon? (Doesn't look like it will, see)
This would go upstream into sshfs-xattr... that's why I am interested.
I've read here that I can't mix SAS and SATA2 drives on the same controller - is that really true? I understand I shouldn't mix on the same channel, but the same card!?
For example, when I plug my 6-drive backplane (IBM x3650-7979) via SFF-8087 cable connector into an Adaptec 2405. Spinning 2 SAS drives (Ch.1), and 4 SATA2 drives (Ch.2+3) - Would it work?
I think this applies to various server hardware. My IBM here has 3 different ways of managing the system with IPMI and related technologies:
1) Standard IPMI connectors which I can connect to with lm-sensors and ipmitool, when logged in via ssh
2) The Baseboard Management Controller (BMC) which uses the same IPMI facilities to tell me about the system status, allow me to reboot. The only plus is that it can be connected on a separate IP address with its own small user management.
3) The Remote Supervisor adapter (RSAII) is like 2) but with the benefit that it is a separate small computer, with own Ethernet and CPU, with the sole purpose of managing the system it lives on.
I understand that each of these technologies represents different levels of "hardware sophistication", with each step up rendering lower steps redundant at times, BUT:
I see the benefit of 3) as it provides lights-out management, but why does 2) expose it's own IP inroad? When the system is ON, I can access IMPI via shell, when it's OFF I can't access BMC and shell. Is the extra IP address and user/login stuff a gimmick??
I'd like to get best-practice advice on how to setup a scheme where 250 users or more have OS independent access to their own data-dir / user-dir on a central server in the local network. from the server's perspective:
Would you set up a big /users dir on the server with subdirs for each user with quota, on one big partition (underlying fat storage, of course)?
Or would you setup a dedicated partition for each user living in a big ZFS, LVM disk pool?
Dirs would be shared over the network via SAMBA, NFS or probably WebDAV, HBA, aoE - not fully knowing what the latter do to the permission case...
What are the pros and cons of "dedicated partition" vs "permission based shared partition"? What's best practice today?
Can someone give me pointers where to look in order to debug why this fresh install of Windows Server 2008 R2 with an untweaked MS SQL-Server R2 which is serving just one ~5GB db to 2 clients crashes once a week?
(All updates applied, no other software running, no other roles. No Hyper-V, running on bare metal. Intel Core i5 660 @ 3,33 Ghz, 16GB RAM, 64bit Windows Server R2)
UPDATE:
I looked into the logs, filtered Windows>System logs for critical and errors only, found (translated into English): ERROR: Service Control Manager; the service "SSPORT" could not be startet, file not found CRITICAL: Kernel-Power
--these are the only severe looking things in this log, possibly unrelated. Nothing in Security logs, and in Application only MSSQL complaining about that it can't connect to the reporting server, right after restart (of which I read in the MS KB that it's normal after restart)
Two software RAID1's,
0: 2x 15K SAS drives, 73GB
1: 2x WD Caviar Green, 3TB
So the first RAID is a faster one, the second a slower data grave. What happens when I add both to a LVM logical volume so they form one large partition?
I don't expect lvm to intelligently move more often used data to the quicker hardware... right? Any suggestions for simple multi-tiered storage with these 4 drives?
Unanswered so far: Will the lv slow down, as in "the weakest link"? Or will it randomly slow down, with some blocks coming from fast, some from slow disks?
There's only one SAS, right? Serial attached SCSI?
The female connectors:
SATA2 looks like: ------------| |-------
And SAS looks : ------------+-+-------
For me, the SATA2 male connector looks like it has little corners in the middle that would prevent it from sliding into SAS, which doesn't have the little gap to allow the corners in. Is this correct?
I've never seen a Linux server I admin. How can I find out if it has IPMI of some sort installed?
I tried answering my own question by installing ipmitool:
Setting up ipmitool (1.8.9-2) ...
Starting IPMI event daemon ipmievdipmievd: using pidfile /var/run/ipmievd.pid0
Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory
Unable to open interface
failed!
invoke-rc.d: initscript ipmievd, action "start" failed.
was the answer. Can I find out what the ISP uses to remote-control the server or how the ISP probably hooked it into their remote-management HTML panel?
One supermicro enclosure comes with "3 SP382-TS 380W Power Modules for a total of 760W".
2x380W = 760W, okay, but there are 3! So where is the missing third 380W equalling to 1140W?? Is it hot-sparing, kicking in when one dies?
Let's assume I build a storage system where each user gets his/her own encrypted filesystem on top of a ZFS pool. Would deduplication on pool-level still work when each fs is encrypted?
What encryption options do I have or which should be favoured? Is anyone using such a setup, where the userspace is opaque to the sys-admin?
The docs state:
"On systems with 1,000s of ZFS file systems, provision 1 GB of extra memory for every 10,000 mounted file systems including snapshots."
What does that mean - how do these people use ZFS? Are they using ZFS inside per-user mounted files? Or does it refer to really huge datacenters??
I always thought ZFS advocates for a single file-system, a giant one. Why do the docs talk about people mounting 1000s??
What about this idea:
...as a kind-of poor-man's replacement for revisiting all code routines using the db and implementing memcached or similar.
On a remote server the boot/system disk is corrupting. It is running Debian 4.0 on an Intel Core 2 CPU 4300 @ 1.80GHz.
Is it possible to "prepare" a bootable disk I'll then send to my ISP to install?
I would like to setup Debian 6 on the disk, copy/install most of the services running on the remote machine on a local test-bed machine, then unplug the boot-disk from the test-bed and send it to the remote-hands of my ISP to install. To minimize downtime.
Problem is, I've got only AMD Athlon based test-machines at hand. I know that at least Windows would install so many architecture dependent stuff on install that such a swap would be nearly impossible. But is Linux "universal" enough to boot once the AMD based install is hooked into the Intel based system??
I've got a work-group of 5 people. All of them have their dedicated machine with Windows 7. They are all connected to a central server running Windows Server 2008 R2, currently mainly offering the file-server role, one shared data directory for all.
Now, what is considered best-practice when these users want to connect to "their desktop" while on the road. Possibly storing everything desktop-related on the server so not all these desktop machines have to be running 24/7?
Some more details:
I could port-forward the remote-desktop port 3389 to different machines via different public ports, but if possible, these machines should be allowed to be powered down. Also, I think, security-wise, exposing all these machines to the WAN is tricky, even on non-standard ports. Only having one machine, the server, handling remote connections sounds better, possibly executing a personalised desktop environment on the server. Possible?
So far, I've tried to separate user data from the users' machines as good as possible, but had little luck. On *nix having a homedir via NFS is simple, in MS world "Remote profiles" seem to be considered unreliable and get bogged up once these "profiles" get big and have to be pulled over the network on each login. mklink and thus warping a local Win7 Profile dir to a server share seems hackish and does only work on the most up to date Win OSs. Also I so far ran into all sorts of permission and architecture problems, for example when I try to use that same Win 7 Profile dir with a similar set-up user on Win Server locally, ouch.
Being on a middle ground right now, I have some data server-side and some-local data. For example people's deskside Outlook pulls files from the server (which is considered unstable by MS, btw, at least as I understood it; having fingers crosses when data eventually gets corrupted).
More hackish middleground: I emulate a server-side Desktop by having the same apps on deskside machines and on the server. People log in to the server only, with a same-username,/ same-password Profile (ouch) on the server and find a desktop similar to theirs on the real deskside system (arg!).
Is Microsoft Terminal Services a solution? So far I couldn't wrap my head around what it actually does. Or how it could help me.
I am faced with a Windows hardware/software problem left over from another person. It's on me to resolve. It's a mission critical setup. The situation is:
I've got a physical server machine with:
-Disk C:\ (one disk) containing a basic install of Windows Server 2008 R2, formerly Win Vista Pro, now gone.
-Disk D:\ (software Raid) containing a VirtualBox disk image of a configured Windows Server 2008 R2 running SQL Server R2 among others.
What shall I do now?
Migrate all the stuff from the configured VM to the basic but
natively installed C:\ Windows Server 2008 R2 (with the possibility
of breaking stuff)? Or,
Setting up the machine to "natively boot" the VM with the help of bcdedit.exe (something I've read about, what I've never done, what I don't know of if it works, if it hits performance, or if it is stable for production)
For me, being old skool, I am in the process of de-virtualising everything (option 1). But I'd be happy if someone suggests I am ok to go down the "natively boot" route.