I'm running 10.04 64-bit diskless on ESXi, installed as a minimal virtual machine. I want this server to access an iSCSI drive. The machine can view the iSCSI shares with iscsiadm, and can even log into the drive. When I do an iSCSI login, this appears in /var/log/messages:
Aug 17 11:08:21 ubuntutest kernel: [1123295.329972] scsi4 : iSCSI Initiator over TCP/IP
So, it appears that open-iSCSI is working correctly. But no new /dev/sd* nodes appear, and nothing new appears in /dev/disk/by-path. I'd expect to see /dev/disk/by-path/ip-XXXXXXXXX. fdisk -l shows nothing but the boot drive. My guess is that the "minimal kernel" doesn't include some necessary module or driver.
I'm putting a server together and have run into a boot up problem. (I thought about putting this in the server forum, but it might be a more generic problem that others have seen and know how to rectify.) The install seems to have gone just fine. I have the /boot partition on an internal IDE drive. The rest of that drive and another are mirrored in a Raid0 configuration (using the Linux software to do that) for data storage. The swap partition is a part of the Raid5 SCSI array that also has the / (root) partition on it.
After installation it would not finish the booting process. I suspected that GRUB didn't like all the Raid arrays and such, but it seems to be fine. I can say that because the machine will boot into rescue mode with the GUI splash screen and I have access to the whole directory tree. I have already searched on-line and following prudent advice, ran the yum update while in the chroot /mnt/sysimage mode. That only took overnight to download and most of this morning to complete. Still no dice. Used vim to delete the rhgb quiet commands in the grub.conf file so I could see where the kernel seems to be hanging.
So right after the "Creating initial device nodes" is a line about my generic PS2 wheel mouse. So I tried a USB mouse. Got more output so tried swapping out to a USB keyboard. Got a little further with more information about input devices, but still stops. Also, I tried a PCI video card just to make sure the onboard video wasn't the problem - no change. So, if someone in the Fedora community knows what loads up or is configured right after the mouse and keyboard, I might be able to figure out what's causing the computer to hang during the boot process.
I'm trying to increase the size of an iSCSI device. On the LUN side, the provision size has been expanded already. I expanded the size of this device from 15GB to 30GB.
I am running RHEL5.5 its a fresh install and we are testing Xen Virtualization. We are wanting to use our iSCSI SAN for the VMs. I have created the initiator iqn, and discovered the target address. We are connected to the target, but there is no new block device in /dev.
I have a server wich is connected to an iSCSI storage and gets harddisks from this storage. Sometimes I have to add new disks to this server. Everytime I add a disk and make an /etc/init.d/iscsi restart on the server the disks don't have the same device name as before the iscsi restart.
It should be possible to gave the disks persistent names using udev rules. Now I tried to create different rules in "/etc/udev/rules.d/99-static-iscsi-names.rules" e.g.
# /dev/sdc KERNEL=="sd*", BUS=="scsi", PROGRAM=="/sbin/scsi_id -g -s %p",RESULT=="360a98000503355344c4a576864467877" NAME=="sdc%n" In "/etc/rc.local" I added "/sbin/start_udev" and in "/etc/scsi_id.config" I added the line "vendor="NETAPP",model="LUN",options=-g"
I have never done this before but stood up a new Debian (testing) x64 system. It only has 146GB available for RAID1 so I created a 500GB iSCSI LUN on my NAS device on the network and am really confused how I can attach my Debian to the iSCSI LUN I created. Right now the OS is installed all locally on the machine but I would like the iSCSI LUN to be the /home directory for mail storage. Is this possible or do I need to mount the LUN to a newly created folder / mount point that is locally attached?
I am planning on setting up a 4x1TB RAID5 with mdadm under Ubuntu 9.10. I tried installing mdadm using "sudo apt-get install mdadm", all worked fine except for the following error: Code: Generating array device nodes... /var/lib/dpkg/info/mdadm.postinst: 170: /dev/MAKEDEV: not found failed. The end result is the /dev/md0 device has not been created, as can be seen here:
Code: windsok@beer:~$ mdadm --detail /dev/md0 mdadm: cannot open /dev/md0: No such file or directory After googling, I found the following bug which describes the issue: [URL] However it was reported way back in April 2009, and it does not look like it will be fixed any time soon, so I was wondering if anyone knows a workaround for this bug, to get me up and running?
fc10 86_64Every time I login and do not run any programs, the swap space used goes from 300 MB to 6GB. Memory stays at about 320 MB (out of 4 GB). I takes several minutes to fill the swap space, and when it gets full, it then goes back to 300 MB. Then the system runs ok.The system monitor shows no processes (other than itself) running, but cpu usage is about 20% on 1 of the 2 cpus
I'm unable to login to my Kubuntu Lucid. The login screen takes my password, blanks, then returns me to the login screen. I'm getting some graphics errors when running from recovery mode as well as the no space left on device error when attempting to start x from the terminal. Here are some outputs: When starting from recovery mode, selecting failsafeX from the Recovery Menu:
[Code]...
I've come to the conclusion that my root partition is full, but I'm not sure how to clear space, or how much to clear once I work out how to do it. I removed a few packages with apt, but it doesn't seem to have any effect. df -h shows that root is 100% full, yet it has 3GB free. I've grown comfortable with Ubuntu in the couple years I've been using it, yet this level of problem-solving is a bit nerve-wracking to me. I've been considering reinstalling (this machine is running Lucid upgraded from Karmic and Jaunty and has a few oddities), but I hate the idea of being forced to reinstall because I can't overcome this problem. If you need any other information or outputs from terminal commands, I'm happy to provide it.
I hope this is not a totally ridiculous question. I use putty often to remotely connect to my laptop at home. I recently learned about the Linux framebuffer, and was curious if there was a tool like putty that would let me remotely connect to my laptop but use the /dev/fb0 device for forwarding graphics. Sort of like a vnc or X11 forwarding, but instead of using X, gdm, kdm etc. we use the framebuffer.
I'm trying to work with a project manager and some hardware guys and all of us are trying to get a grasp on the UEC.
1. Can a VM span multiple nodes?
2. If a website scales beyond 100% CPU on a node, does it automatically spill over to another node and start using processing power there?
3. Do I have to rewrite the code of my web app (such as WordPress) so that it works in a private cloud?
4. Does KVM support virtual memory such that when I run out of RAM it uses paging?
5. How is UEC any different than me bringing up several servers with several VMs inside, and then moving VMs around as I need? Is it really a cloud where the VMs exist in a cloud, not necessarily a node?
6. If I power off a node while it's hot, how can I assure myself that the data will be okay?
7. Do I want to use SCSI attached RAID, or NAS, or NFS? We're simply trying to host thousands of blogs and LAMP sites in a way that we can handle spike loads.
8. Do I attach the disk volumes to the cluster controller and then run a command such that the nodes can mount those virtual volumes?
9. Do I need to run another command on the VMs to mount volumes that the nodes can see?
10. What is the role of the storage controller and how is it separate from the cluster controller?
11. How many systems, minimum, do we need to get this tested out?
12. For testing purposes, can I install UEC on an Intel P4 server and then a node on Intel P3 server -- even with some limitations?
13. Let's imagine I have 5 node servers, and one catches fire and is a total loss. Can I get those VMs back and get that data back? What would have to be the proper config to make that happen?
I want to connect two systems in clustering concept. am new for clustering configuration. I have installed ubuntu 9.10 server edition in two system. what do to the next step to configure clustering in ubuntu 9.10 server edition.
How to set the anchor node in NS2? How to define the channel propagation model as Log Normal Shadowing instead of two ray ground reflection model in NS2?
In the comment line, I've got the parent node of the current node. Parent node is "User" node. I would like to obtain the children of this parent node, say "name","surname " etc. How can I do this? Is there any function such as : "$parent_nodes->bringChildren('name')".
i need to discover all network nodes and their ips is their a way to do it, then if i need to discover certain nodes which runs certain program. I'm using Fedora14 and FreeBSD
I have a cluster script (CentOS 5.4) and I need a UPS for a shutdown. It should look like this ... The first node must wait until the second node has completed several actions and has been completely shut down. So I need a script with a shutdown wait until the other node is ... I was thinking of a test using ping. Alternatively, I would also have an IPMI interface. However, the "pause" may last a maximum of 10 minutes ... Then the wait must be ended.
Now that Ubuntu is at a LTS stage with Lucid and is easier to install PV in XenServer, I have been wanting to drop my current debian VM nodes in favor of Ubuntu 10.04. So far I have DRBD8 and OCFS2 all working nicely as primary/primary and I am on the next stage of my setup by installing ldirectord and heartbeat. Can anyone shed some light on heartbeat-gui (hb_gui) as it no longer seems to be in the repositories (since Jaunty)?[URL].. I understand that Lucid now uses pacemaker as part of a clustering suite but I have no idea if I really need it. I really liked being able to run hb_gui and logging into the heartbeat cluster to manage resources via point and click so if anyone has any hints on replacements or newer preferred methods to accomplish the same tasks via point-n-click,
I am setting up a computing cluster in my lab, as below. all the "eth0" IP addresses are static (for cluster communication) and the "eth1" of the front node is the only one connected to the internet through lab's DHCP server (which is connected to a centralized computer center in the university). The thing I wish to do is to do some sort of IP masquerading to enable all the nodes to have internet access. I actually google around and read some books. The similar things I came across is setting rules in iptables but I did not manage to get any of them working. I am using Ubuntu Lucid 64-bit on all machine.
Are there any prerequisites which much be met in order for udev to create /dev/usb/hiddev* nodes?
I have two Slack 13.0 systems, each connected to a APC UPS via USB, one of which creates a hiddev0 node and the other doesn't. apcupsd can't find the UPS unless this node is present.
The "working" system has been stable in its current configuration for so long that I honestly can't remember if I had to tweak anything in this regard. I've verified that it doesn't use any local udev rules to create this node, and to the best of my knowledge there are no external scripts or processes which would create it either.
The non-working system has a similar configuration - no local rules and/or scripts which touch usb devices or nodes.
I have an annoying problem with a 2 nodes cluster that I don't understand.
One of my server is totally shutdown.
When I start cman after a reboot on the other server, it hangs on "Starting fencing...". The process which is actually hanging is : /sbin/fence_tool -w -t 300 -m 45 join
When I start this program manually, I get this output :
# /sbin/fence_tool -w -t 300 -m 45 join Waiting for all 2 nodes to be members Waiting for all 2 nodes to be members Waiting for all 2 nodes to be members Waiting for all 2 nodes to be members (and so on...)
What I don't understand is why the fence_tool program wait for the 2nd node to join the cluster and don't fence the other node.
I have an iSCSI target on my Ubuntu 10.04 LTS server that is giving me access to 3 x 2TB volumes. I am using LVM to create 1 big volume and using it as a place for my backup software to write to. Upon initial setup I tested to see if the connection would still be there after a reboot, and it was. After going through the trouble of configuring my backup software (ZManda) and doing a few very large backups I had to reboot my SAN (OpenFiler) for hardware reasons. I took my server down while performing the maintenance and brought it back up only after the SAN work was done. Now, the LVM volume is listed as "not found" when booting.
Using iscsiadm I can see the target but LVM doesn't know of any volumes that exist using those targets. I need to get it back up and running and then troubleshoot the reboot issue.
I'm having a problem with NFS. I'm trying to mount /home, which is on the server, to the client nodes. The server is currently running FC14 (64 bit) and I'm in the process of updating all the clients nodes to the same OS. Currently, all the client nodes are running FC8 (32 bit) and are capable of mounting /home on the server. However, the one client node that I have updated to FC14 is having problems mounting /home. I can ssh to the node with no problem however I get the following message:
Could not chdir to home directory /home/myuserid: Permission denied. Now, if I excute the following command: "sudo mount -a" I get no errors and now I can see my home directory.
I am deciding to whether use OSPF or RIPv2 on my network. The key point for me is how good each of them are in dynamically adding and removing nodes. I am pretty sure that OSPF works well if a node is suddenly dropped or a new one is introduced( Please correct me if I am wrong). how good RIPv2 is at this.
I use Slackware 13.0. I just upgraded from 2.6.32 to 2.6.35. On 2.6.35, the kernel supplying strange non-existant node like hda* and many symlink like cdr cdr0 cdrom cdrom0 etc. Well, i use libata and disable the old ATA driver, so my HDD should only be displayed as sd**. Also, i've checked all the udev rules, and i don't see any SYMLINK entries that would produce that cdr cdr0 links. My 2.6.32 and 2.6.35 config are identical. Everything were working normally on 2.6.32. Btw, i've tried to upgrade to 13.1. Then boot the standard huge.smp kernel. The same problem occured too. And with other problem i get with 13.1, i decided to downgrade to 13.0.
As I was updating machines today, I was wondering... Is there some way to cache the RPM files locally that yum is grabbing? Like many, I have a number of VMs I run, and it most updates are the same between the nodes. I'm guessing I'd need some sort of http cache? Just looking for pointers, not sure what I'm looking for, however.
