April 2013 - luv-main

zram
by Russell Coker 24 Apr '13

24 Apr '13

There has been some discussion on luv-talk (which should have been on luv- main) about zram, a system that uses a compressed RAM disk for swap on Linux. I've enabled zram on one of my systems. I ran "free" on all the systems I run and found that only one had any significant swap use, that's a workstation with 8G of RAM and 512M of swap in use. I didn't bother enabling it on any other systems. I added "zram" to /etc/modules (modules that are loaded on boot on Debian). I created /etc/modprobe.d/zram.conf with the following contents: options zram num_devices=2 This gives two devices and it seems that you are expected to have one zram device for each CPU core. I'm not sure if this is really required but it doesn't seem to do any harm. To enable zram on boot I added the following to /etc/rc.local: echo 1073741824 > /sys/block/zram0/disksize echo 1073741824 > /sys/block/zram1/disksize mkswap /dev/zram0 mkswap /dev/zram1 swapon -p 100 /dev/zram0 swapon -p 100 /dev/zram1 The above sets each of the two devices to 1G of storage and enables them as swap space. I haven't yet rebooted the system to make sure that there are no errors in the above, but as it's what I types on the command-line it should work. Also I haven't tested zram extensively. Finally I'm running the Debian/Experimental 3.8 kernel. I don't know how it would work on other kernels. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

1 0

Re: [luv-main] Redundant network links.
by Russell Coker 23 Apr '13

23 Apr '13

Firstly I'm moving this to luv-main because it's about serious Linux networking issues. On Sun, 21 Apr 2013, Tony Langdon <vk3jed(a)gmail.com> wrote: > About to rebuild the network here after moving house. One issue I'm > contemplating is performance and redundancy. One part of the network > can't (easily) be reached by Cat 5/6, running cables t that part of the > house would be too messy at best. I have two ways I can bridge this gap > - Powerline Ethernet adapters, which have worked extremely well in the > past, or WiFi, using an access point in client bridge mode. > > Now the powerline adapters do work extremely well, with a rated speed of > 85 Mbps. I've never had an issue, except for the switchmode supply of > one laptop, which trashed the link (took a bit of detective work that > one!). The biggest weakness of these devices is that they can't be > battery backed up. If the mains goes down, so do they. http://etbe.coker.com.au/2010/08/04/clusters-dont-work/ I wrote the above blog post about my experience with clusters. Some of this applies to failover of links. One significant difference between a home network and a serious server network is that most of the functions of your home network don't matter much when you are asleep or away. Therefore a redundancy which involves you logging in as root and running a route command will work a lot better on a home network. Of course my experience is that having a sysadmin login as root and manually fail things over is better than any cluster software implementation I've seen, but that's another issue. > WiFi can be backed up, especially since a lot of the gear I have will > happily run off a 12V battery, and some of the systems on the far end > will be running off a battery backed DC supply. However, performance > with the WiFi solution isn't as good. > > Is there a way I can (easily and cheaply) arrange to run on the > powerline devices by default and fail over to WiFi, if the power goes down? How often does power go down anyway? -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

5 9

debian "experimental" kernels
by James Harper 21 Apr '13

21 Apr '13

Is anyone using kernels from Debian experimental? I needed something newer than 3.2.x (testing ceph but I'm getting kbyte/second write performance!) so I tried a 3.8.x kernel from Debian experimental but it just reboots with no indication of why (using serial console via bmc) James

6 20

Invitation to connect on LinkedIn
by Peter Drake 21 Apr '13

21 Apr '13

LinkedIn ------------ Linux, I'd like to include you in my network to share updates and stay in touch. - Peter Peter Drake Teacher at Victoria University Melbourne Area, Australia Confirm that you know Peter Drake: https://www.linkedin.com/e/-yhvhf6-hfqhmow1-y/isd/12623853746/illMn13Y/?hs=… -- You are receiving Invitation to Connect emails. Click to unsubscribe: http://www.linkedin.com/e/-yhvhf6-hfqhmow1-y/AsY5QCVyLGMDpk6UevYY8mLJhvdwey… (c) 2012 LinkedIn Corporation. 2029 Stierlin Ct, Mountain View, CA 94043, USA.

5 6

Video problems - NVidia/Nouveau
by Jason White 19 Apr '13

19 Apr '13

My desktop workstation has been encountering intermittent video issues, including illegible screen displays and cases in which the monitor doesn't respond at all after Linux is loaded. Rebooting the system always seems to fix it however. This isn't so much of a problem for me, as I have a vision impairment and use braille and speech output, but it is an issue when someone else wants to use the machine. The only relevant error I found in the kernel log is: Mar 20 15:05:58 jdc kernel: [23207.760016] [drm] nouveau 0000:40:00.0: DDC responded, but no EDID for DVI-I-2 (after turning the monitor off then back on, after which it still didn't display any video.) The video card in this machine has been replaced twice over the years, both times under warranty/service contract, and with the same model of card, so I have my suspicions. It's a very up to date Debian Sid system, kernel 3.2.0-4-amd64 #1 SMP Debian 3.2.39-2 x86_64 GNU/Linux I didn't find any relevant Debian bug reports. Obviously I want to know whether it's hardware or software, and if hardware, the card itself or something else.

6 12

RAID, again
by James Harper 17 Apr '13

17 Apr '13

Now I have another RAID problem. Each disk is partitioned like: /dev/sd[abcd]1 - bios boot partition /dev/sd[abcd]2 - md0 - /boot (I believe grub can now see inside lvm but this config predates that) /dev/sdX3 - md1 - lvm - /root + others I have been progressively replacing disks. sdd was fine, sdc was fine, then I replaced sda to test that booting worked (changed bios to boot from sdb, added new sda, added to raid, etc) and that didn't work because I mucked up the gpt bios boot partition. Having fixed that, I rebooted to test booting again while md1 was still rebuilding onto sda3, and it drops into the initramfs shell on boot because mdadm first sees sda, then tries to add sd[bcd]3 but because each of those disks say that sda3 is not consistent the rebuild fails leaving just sda3 present as a 'spare' drive. In the initramfs console I just manually stop the array, add sd[bcd]3, then sda3, and it's all good again, but I didn't think this was the way it was supposed to work. On the most recent boot I zero'd the superblock of sda3 before re-adding it, thinking maybe it had picked up some bitrot somewhere along the way, but I'm not confident. Because the server is responsible for running backups and last night's backups didn't work because it was down, I'm not going to touch it again until tonight's backups are complete, but any hints on how to make this work smoother next time would be appreciated! And just to recap, rebooting a RAID comprising of /dev/sd[abcd]3, when /dev/sda3 is being rebuilt, results in a boot that drops into the initramfs shell because it appears that mdadm tries to add /dev/sda3 first then rejects the other 3 disks because they say /dev/sda3 is inconsistent (which it is). Thanks James

3 5

spamassassin and detailed logs
by Russell Coker 17 Apr '13

17 Apr '13

spamd: result: . 1 - DKIM_ADSP_DISCARD,SPF_PASS,T_RP_MATCHES_RCVD,UNPARSEABLE_RELAY Currently I'm getting log entries like the above from SpamAssassin. spamd: result: . 1 - DKIM_ADSP_DISCARD=1,SPF_PASS=-1,T_RP_MATCHES_RCVD=1,UNPARSEABLE_RELAY=0 I want to see something like the above with numbers for each test so I can easily determine why the total was determined. I know that I can search through the SA configuration for each test, but this is time consuming and has the problem that the SA configuration might not do what I want. add_header all Status "_YESNO_, score=_SCORE_ required=_REQD_ tests=_TESTSSCORES_ autolearn=_AUTOLEARN_ version=_VERSION_" According to the below URL adding the above to local.cf should give the result I want. But on Debian/Wheezy it doesn't. Can anyone give me a pointer? As an aside I'm configuring my personal server on Debian/Wheezy and when I get it right I'll do it on a bigger server running CentOS 5. So if there are any relevant differences between Debian and CentOS in this regard I'd like to know about them. http://mail-archives.apache.org/mod_mbox/spamassassin- users/200703.mbox/%3C20070304222615.D13968(a)duane.dbq.yournetplus.com%3E Thanks. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

2 6

ups <-> psu comms
by James Harper 17 Apr '13

17 Apr '13

Is there such a thing as a ups to power supply communications channel over the 240V lines, such that the UPS can talk to each power supply in a server and figure out what is plugged in where? It seems such an obviously useful idea but I can't see that such a thing exists anywhere. Comms would only need to be very basic (hundreds of bps would be sufficient) - PSU would tell the UPS what its maximum rating was and maybe an identifier (eg for grouping redundant PSUs), and the UPS would tell the PSU when it was on battery etc. I'm shopping for UPS's and looking at features etc. I'm opting for network based comms on my UPS's these days but even seems like a lot more mucking around than it could be. James

4 5

file sharing over the Internet
by Russell Coker 15 Apr '13

15 Apr '13

A client has a Samba server running ZFS that's working quite well for what it does. The problem is that a bunch of staff want to work from home and have been using Dropbox and Gmail to share the files - not what we want. Among other things we'd like to have the files stay within Australian jurisdiction. What's a good way for sharing files to Macs and Windows systems over the Internet that's easy to use and relatively secure? -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

13 17

Re: ZFS vs RAID (was gpt and grub)
by Petros 15 Apr '13

15 Apr '13

Quoting "Craig Sanders" <cas(a)taz.net.au> > On Fri, Apr 12, 2013 at 03:31:20PM +1000, Kevin wrote: >> On Fri, Apr 12, 2013 at 3:17 PM, James Harper >> <james.harper(a)bendigoit.com.au> wrote: >> >> > My last remaining reservation on going ahead with some testing is is >> > there an equivalent of clvm for zfs? Or is that the right >> approach for zfs? >> > My main server cluster is: >> > >> > 2 machines each running 2 x 2TB disks with DRBD with the primary >> exporting the whole disk as an iSCSI volume >> > 2 machines each importing the iSCSI volume running lvm (clvm) on >> top, and using the lv's as backing stores for xen VM's. > >> > How would this best be done using zfs? > > short answer: zfs doesn't do that. > > in theory you could export each disk individually with iscsi and build > ZFS pools (two mirrored pools). if that actually worked, you'd have to > do a lot of manual stuffing around to make sure that the pools were only > in use on one machine at a time, and more drudgery to handle fail-over > events. seems like a fragile PITA and not worth the bother, even if it > could be made to work. James, you want ZFS mirrored over the net as the backend for VM storage (virtual disks?) In case you really want.. "Poor man's version" similar as Russell described (zfs send/receive) is used by me: I am looking for running Jails and VirtualBoxes (well, the later is just one here..) and run zfs snapshot/send/receive to send all filesystems of the active Jails/VMs to the other side. In case one machine is broken I bring them up on the "other side". To mirror them without delay, you can create, one both sides, a volume on top of ZFS (zfs create -V 5gb MyPool/MyVM), export them over iscsi and mirror them. The fail-over between Virtual Machines is a problem independent of the sharing on the file system layer. A semaphore on the shared volume can be used for a cluster solution to manage it. BTW: You can put any file system, including a clustered one, inside these ZFS volumes, and combine the benefits of ZFS volume management with the benefits of the filesystem "inside". Regards Peter

1 0