|
zorachus posted:The guy from the build team swears that he didn't touch the LUNs on the repurposed server, but it's possible that their automated new build process reinitializes any attached storage. I just thought that it was strange that I was able to recreate the devices with mknod and that mount recognized them as ASM partitions. If the re purposed server used ASM, it might have recognized them as ASM and automatically zapped them. I'm still a newb at the oracle part of oracle, tho i've been supporting OS operations for oracle boxes for years now. Did your DBA get to the part where he imported the devices into oracle upon setup? From what I remember, that would have done it. From what i remember w/ powerpath, you shouldn't need to mknod powerpath devices, EMC powerpath should be doing this for you. I don't work with it THAT much, but perhaps there's some powerpath magic you could have done to get the nodes back. Having said that, with what little info you have, i'm pretty sure the 3rd re purposed server zapped the contents of the storage.
|
| # ? Sep 26, 2011 14:07 |
|
|
|
| # ? May 18, 2013 19:41 |
|
|
Corvettefisher posted:Anyone know a simple fix to get Squid to cache youtube videos? One of the main reasons I have squid is for that purpose Good luck. Here's some solutions, but none are perfect. http://wiki.squid-cache.org/ConfigE...Content/YouTube Back a long time ago (when I used to use squid) I was able to just increase the max_object size to something like 2GB and that worked great. But then the server stores a copy of every video watched so that uses a ton of disk space (hurr the point of a cache right?) Then they merged with Google video so not all requests come from YouTube, and then they made it so you can stream a video from more than one server, it all turned into a huge pain in the dick.
|
| # ? Sep 26, 2011 14:13 |
|
|
Thanks to everyone who replied! I will get started on doing those things, having some direction already makes me feel a lot less lost. Bob Morales: Thanks for the offer, I really appreciate it. I have a server in my basement running a couple of VMs that I've been using, plus I work for a hosting company, and can set up test accounts if I want, so I'm okay on that front.
|
| # ? Sep 26, 2011 14:15 |
|
|
Bob Morales posted:Good luck. Here's some solutions, but none are perfect. Yeah and all the different resolutions... sigh welp something is better than noting I guess
|
| # ? Sep 26, 2011 14:16 |
|
|
nitrogen posted:If the re purposed server used ASM, it might have recognized them as ASM and automatically zapped them. I'm still a newb at the oracle part of oracle, tho i've been supporting OS operations for oracle boxes for years now. I think that you're probably right. I'll talk to the build guy and see if he'll let me take a look at their build process. It's worth mentioning that the repurposed server was turned over to the DBAs already, so even if the build team didn't touch those LUNs, the DBAs may have. Just scanning for ASM disks shouldn't affect them, but it's looking more and more like they reinitialized them. I agree that you shouldn't have to mknod PowerPath devices; it was a last-ditch effort to recover the partitions. Thanks for the insight. We assumed that they were gone, but I had to at least explore every avenue available.
|
| # ? Sep 26, 2011 14:24 |
|
|
Corvettefisher posted:Yeah and all the different resolutions... sigh welp something is better than noting I guess acl youtube_domains dstdomain .youtube.com .googlevideo.com .ytimg.com http_access deny youtube_domains 
|
| # ? Sep 26, 2011 14:27 |
|
|
Bob Morales posted:acl youtube_domains dstdomain .youtube.com .googlevideo.com .ytimg.com BUT HOW DO YOU SUPPOSE A BUSINESS RUNS WITHOUT 30_minutes_of_Cuite_kittens_1080p.flv Seriously managers need this stuff for... uhh... IMPORTANT THINGS! or so I am told
|
| # ? Sep 26, 2011 15:00 |
|
|
Corvettefisher posted:BUT HOW DO YOU SUPPOSE A BUSINESS RUNS WITHOUT 30_minutes_of_Cuite_kittens_1080p.flv Seriously managers need this stuff for... uhh... IMPORTANT THINGS! We spent a bunch of time making a Youtube account, making the page look pretty, getting the video links working with our CMS, etc Then we start getting the question of "We can't see your videos because YouTube is blocked by our IT department" (We have demo and marketing videos of our products that we show on our website) So now we had to track down some Flash player and host the videos ourselves on our overtaxed webserver. Then re-do all our pages. Thankfully it's just bandwidth which we have 150mbs unlimited of. edit: ^^ code:Bob Morales fucked around with this message at Sep 26, 2011 around 15:52 |
| # ? Sep 26, 2011 15:50 |
|
|
I just did a dist upgrade on a debian box from squeeze to testing. mysql will no longer start >:(. As far as I can tell, debian has changed from using /var/run to just /run? So /var/run/mysqld is now /run/mysqld. Fine, I did a find/replace in /etc/mysql/my.cnf and debian.cnf. Still no dice. /var/run/mysqld and /run/mysqld both exist, exact same permissions/user/group and both empty. What else am I missing here? syslog just has an error that it can't find the socket in /run/mysqld. ps I don't really know what I am doing, this is just a box I gently caress around on (gently caress things up on). Kaluza-Klein fucked around with this message at Sep 27, 2011 around 02:13 |
| # ? Sep 27, 2011 02:08 |
|
|
Check your /var/log/mysql* logs
|
| # ? Sep 27, 2011 07:08 |
|
|
spankmeister posted:Check your /var/log/mysql* logs Even after turning logging on, the mysql log was empty. syslog was saying that it couldn't create/find the socket file. On a restart of the server, mysql came up, but eth0 didn't. Turns out the server is all sorts of confused, as it was set to use dhcp, but didn't have the ifupdown package or a dhcp client installed. After sorting that out, things seem to be working ok, for now. I am really thinking I need to back up everything and  this sucker. It seems to have problems all the time, and I swear not all of them can be pinned directly on me!
|
| # ? Sep 27, 2011 12:47 |
|
|
I have a Ruby On Rails related question regarding Apache web server, load balancers, and network file systems. Right now we have a web server that runs 15 or so web apps. They're all similar but different, newer, older, whatever. There's also a MySQL database server, both systems run in RHEL and are on a dual-quad Xeon L5520 with 24GB of RAM running VMware. Everything has been fine up until the last few weeks where we launched two new apps and have seen a ton of new traffic. All 8 vCPU's on the web server get pegged from 10:00am to 1:00pm. We're speeding up some pages in the app but the bottom line is we're going to need more capability, if not now in the very near future. The load average on the server goes over 50, it gets really unresponsive over 20. Pages take 5-10 seconds to load if at all. What I was thinking about doing was getting a second web server. I was thinking another dual-quad Xeon with 16 cores, but possibly bare metal. Because of how our site works, I can't simply clone the web server to the other, or deploy to two servers with Capistrano. We need to write uploaded files (and existing files) on the server. I guess we could have some sort of rsync job running between the two, because it's mostly avatar-type images that get uploaded, and app-specific files (custom pages for certain customers, etc). What I was thinking of doing is having some sort of shared storage. I just don't know what kind. Should I have a VM running openfiler or something and share it to both of the web servers? NFS, SMB, iSCSI...? Of course we'd set the firewall to only accept connections from the webservers (the second server will be in the same datacenter but not necessarily in the same rack). Or, should I just share the /var/www/apps directory from the first web server (or the second one?) I could also spin up a VM on our existing server (I have 4 unused vCPU's) and can double the RAM to 48GB (our DB server could use a boost) The end result being we have our current 8-core web server along with a new 16-core web server serving up pages. I was thinking of using HAproxy on one of the servers (probably the 16-core) as a load balancer. Thoughts?
|
| # ? Sep 28, 2011 02:09 |
|
|
Bob Morales posted:Thoughts? Couple dumb questions first off: Is your vmware host cpu-bound/oversubscribed? Have you checked the host and guest? does your guest look like it's io bound or cpu bound? Reason I ask is because this sounds like a classic case of oversubscription of CPU resources on vmware host. PRO TIP: vcpu's on vmware guests don't work like you'd think they would. if you're cpu tick starved, having 2 or 4 vcpu's will make your guest WORSE off. Each of those VCPU's will need a tick before any work gets done. Based on what info you gave, this might be a possibility. I would see stuff like this all the time in my previous position. As far as scaling, scaling out is almost always better for this. nitrogen fucked around with this message at Sep 28, 2011 around 03:55 |
| # ? Sep 28, 2011 03:52 |
|
|
Bob Morales posted:I have a Ruby On Rails related question regarding Apache web server, load balancers, and network file systems. Too many variables there to work with, you should try and figure out where your major bottleneck is. If your DB is slow, then your web app can take a hit because of it. The queries are waiting longer and it keeps processes open longer using more CPU. If your disk I/O is an issue then throwing CPU cores at it doesn't help. Usually NFS would be done with a direct connection, I don't know how well that would work if you have it mounted going through a public IP. You can sync up servers automatically using rsync and inotifyd or lsynd, and then mount your uploads directory via NFS. That is what I set up for a customer the other day and it sorted out their issues with their heartbeat lvs setup. Or just ditch apache altogether, switch to nginx+ rails, get some caching going on (something like memcache, or put varnish cache in front). You have to convert htaccess rewrites to nginx configs, or you do a setup like put nginx in front tell it to serve all static content and send everything else to apache which listens on another port (like 8080).
|
| # ? Sep 28, 2011 07:33 |
|
|
For configuration I would also suggest using some kind of version control. Even just doing 'git init' and keeping it local can be incredibly helpful. Have something that occurred a string of edits ago? Revert that bitch. Totally blew everything the gently caress up and holy poo poo everything's on fire? Rollback son. Learn it, live it, love it. Wisdom.
|
| # ? Sep 28, 2011 08:27 |
|
|
Wheelchair Stunts posted:For configuration I would also suggest using some kind of version control. Even just doing 'git init' and keeping it local can be incredibly helpful. Have something that occurred a string of edits ago? Revert that bitch. Totally blew everything the gently caress up and holy poo poo everything's on fire? Rollback son. This. It just works really well. I learned this from a crusty old Unix admin, he uses RCS for the purpose, which is ancient but still works of course.
|
| # ? Sep 28, 2011 08:32 |
|
|
nitrogen posted:Couple dumb questions first off: We are CPU bound. We don't max the SQL server's CPU's out, we have 150mb/s bandwidth that we are using a total of 6mb/s of, and we aren't swapping. We have 300 clients in the queue at our busiest times according to passenger-stats. We either need more processors, or faster processors. All 8 cpu's are maxed out at our peak hours. We went from 5 vCPU's to 8 yesterday, and it helped, but it's still not enough. I've thought about leaving the busiest 2 sites (the new ones that have been giving us the problems) on the existing server and then spinning up another 4 vCPU (what we have un-allocated) to serve the rest (since we had 5 doing those before with no issues).
|
| # ? Sep 28, 2011 12:33 |
|
|
If I mount a Windows share using CIFS via fstab, will it hang the system on boot indefinitely if the share is not accessible, or is there a way to set a timeout for connecting to CIFS shares? If the share gets disconnected after the system is running, will it reconnect to the share automatically? I could set up a cron job that runs every hour to mount the shares, but I don't want to have to mess with the sudoers file if I don't have to. This system is running Ubuntu 10.04 LTS.
|
| # ? Oct 1, 2011 16:57 |
|
|
Megiddo posted:I could set up a cron job that runs every hour to mount the shares, but I don't want to have to mess with the sudoers file if I don't have to.
|
| # ? Oct 1, 2011 20:28 |
|
|
I am trying to understand chmod better. I understand that there is user, group, others, and all. I understand that the different numbers mean different permissions for these different users. My questions are:
Thank you very much 
|
| # ? Oct 2, 2011 03:42 |
|
|
Crush posted:I am trying to understand chmod better. I understand that there is user, group, others, and all. I understand that the different numbers mean different permissions for these different users. The user is the owner of the file. By default a file you create is owned by you, but you can use the chown command to change that. Same with the group, by default your default group is the group of a file, but that can be changed with the chgrp command. Anybody that is a member of the group has the middle (or 3rd, if 4 digits) permissions. Others is the same as all. The 4th digit of chmod is for "special permissions" http://en.wikipedia.org/wiki/Chmod#Special_modes 4 is setuid, 2 is setgid, and 1 is sticky bit.
|
| # ? Oct 2, 2011 04:14 |
|
|
spankmeister posted:Why would you mess with the sudoers file if cron runs as root?
|
| # ? Oct 2, 2011 04:44 |
|
|
FISHMANPET posted:The user is the owner of the file. By default a file you create is owned by you, but you can use the chown command to change that. Same with the group, by default your default group is the group of a file, but that can be changed with the chgrp command. Anybody that is a member of the group has the middle (or 3rd, if 4 digits) permissions. Others is the same as all. Thanks!
|
| # ? Oct 2, 2011 05:04 |
|
|
Megiddo posted:I was trying to have the mount commands as both a script that runs when a user logged on and a script running at a specific interval. I guess I was confusing the former with the latter and didn't realize that the cron job wouldn't need a sudo command. In cron commands run as root by default but you can run commands as other users in three ways: In the system crontab (/etc/crontab) you put the command like so: */5 * * * * <user> <command> (This command runs every five minutes) This works the same for the /etc/cron.d/ files because they get included to the system crontab. You can edit the respective users' crontab: crontab -u <user> -e */5 * * * * <command> You leave out the user it has to run as Lastly, something I see on old unices: in root's crontab (run crontab -e as root) */5 * * * * su - <username> -c "<command>" This uses root's crontab to run the command as root but it actually runs it as the other user via su. As for your original problem, cifs itself doesn't really have any options to fix that afaik. You could look into autofs/automounter.
|
| # ? Oct 2, 2011 09:58 |
|
|
 ^^^^Megiddo posted:I was trying to have the mount commands as both a script that runs when a user logged on and a script running at a specific interval. I guess I was confusing the former with the latter and didn't realize that the cron job wouldn't need a sudo command. Look into using the Automounter. EDIT: BEATEN. (shoulda looked.)
|
| # ? Oct 2, 2011 14:31 |
|
|
AutoFS seems to work, so thanks everybody for the suggestion. I wasn't sure if it would work with the audio software I'm using and its dropbox system, but so far so good.
|
| # ? Oct 2, 2011 19:49 |
|
|
Megiddo posted:AutoFS seems to work, so thanks everybody for the suggestion. I wasn't sure if it would work with the audio software I'm using and its dropbox system, but so far so good. A program doesn't have to be compatible with autofs, if it can use a file system, it an use autofs. When the program makes a call to an automounted file system, the automounter just mounts it if it needs to.
|
| # ? Oct 3, 2011 03:11 |
|
|
FISHMANPET posted:A program doesn't have to be compatible with autofs, if it can use a file system, it an use autofs. When the program makes a call to an automounted file system, the automounter just mounts it if it needs to.
|
| # ? Oct 3, 2011 04:21 |
|
|
These Fedora cheatcubes are kind of neat: http://fedoraproject.org/wiki/File:..._Cube_Users.png 
|
| # ? Oct 3, 2011 13:53 |
|
|
Any suggestions for a syslog web frontend?
|
| # ? Oct 4, 2011 16:23 |
|
|
bimmian posted:Any suggestions for a syslog web frontend? Try LogZilla (formerly php-syslog-ng): http://www.logzilla.pro/ Splunk is also popular, though I think it's a bit more of a complete package as opposed to just a frontend -- a bit more "enterprise" for better or worse... http://www.splunk.com/download
|
| # ? Oct 4, 2011 17:02 |
|
|
muskrat posted:Try LogZilla (formerly php-syslog-ng): http://www.logzilla.pro/ Too bad logzillas free version only supports 10 hosts. I needed more than that, ended up installing Adiscon LogAnalyzer (http://loganalyzer.adiscon.com/). Works well, good enough feature set for what we need it for.
|
| # ? Oct 4, 2011 20:43 |
|
|
I'm trying to find the best way to merge some simple text files between two computers on my home network. Basically I want to be able to edit the same file on either computer, and be able to switch computers at some point and be able to edit the same file (after the changes). This is simple when I am connected to the network, I can just use tramp in emacs to edit the remote file. But when I want to leave the network with my laptop, I'd like the changes to be applied to my home pc at some point without me manually initiating it. I'd like to avoid dropbox for various reasons, and although git is overkill it would make it easy to handle diffs and I can just set up a cron job to periodically push the changes, but it seems a bit convoluted. rsync seems the best, but the problem is I can't figure out a way to merge the two files if I've edited them both independently when cron runs and wants to copy it one way or the other. Any ideas?
|
| # ? Oct 5, 2011 00:41 |
|
|
Without Pants posted:I'm trying to find the best way to merge some simple text files between two computers on my home network. Basically I want to be able to edit the same file on either computer, and be able to switch computers at some point and be able to edit the same file (after the changes). This is simple when I am connected to the network, I can just use tramp in emacs to edit the remote file. But when I want to leave the network with my laptop, I'd like the changes to be applied to my home pc at some point without me manually initiating it. Have you considered Ubuntu One? Its pretty slick for what you're describing, although I'm not sure how it handles conflicts.
|
| # ? Oct 5, 2011 00:48 |
|
|
Without Pants posted:I'm trying to find the best way to merge some simple text files between two computers on my home network. Basically I want to be able to edit the same file on either computer, and be able to switch computers at some point and be able to edit the same file (after the changes). This is simple when I am connected to the network, I can just use tramp in emacs to edit the remote file. But when I want to leave the network with my laptop, I'd like the changes to be applied to my home pc at some point without me manually initiating it. Have you looked at Unison? (http://www.cis.upenn.edu/~bcpierce/unison/) No personal experience with it, but the description seems spot on for your needs.
|
| # ? Oct 5, 2011 00:51 |
|
|
Thanks guys, I haven't checked out either of those.
|
| # ? Oct 5, 2011 01:14 |
|
|
Ripping my hair out on a new Pacemaker/cman/(OCFS2|GFS) config. I can't use the Pacemaker integration with OCFS2, since the Fedora team, in their wisdom, decided to remove the .pcmk. Supposedly, I don't need to configure fencing in /etc/cluster/cluster.conf (the cman part, of course), and Pacemaker manages things fine if I have o2cb's cluster personality set to o2cb (rather than pcmk, which is unusable, or cman), but I lose fcntl() that way, whereas cman supports it. o2cb/OCFS think they're running fine under cman, and the control daemons are brought up properly, but I'm unable to mount anything (which works fine with the volumes/o2cb tuned to o2cb as the cluster manager). OCFS2 volumes and GFS2 volumes (just to test them) both bomb out with something akin to: dlm_join_lockspace no fence domain STONITH is configured in Pacemaker, fenced is running (with no errors in the logfile there), and everything seems peachy. Except I can't get cluster filesystems to mount. Has anybody dealt with this before?
|
| # ? Oct 5, 2011 20:21 |
|
|
muskrat posted:Splunk is also popular, though I think it's a bit more of a complete package as opposed to just a frontend -- a bit more "enterprise" for better or worse... http://www.splunk.com/download
|
| # ? Oct 6, 2011 13:28 |
|
|
Does any of you guys read the Linux magazines? I usually check them out at Barnes and Noble. Linux Format, Linux Pro, Linux Magazine, and Linux Journal. The ones with DVD's are pretty drat expensive (plus they are from Europe) like $16. I also don't see the point in buying a DVD since we've had cable internet for the last 10 years, so why pay extra for a distro that's a couple months old? Maybe if I only had internet when I could find an open wifi station, and I could use one of those DVD's to install Linux on my laptop. Anyway, sometimes I get some good tips or find out about apps I don't hear about online.
|
| # ? Oct 6, 2011 19:03 |
|
|
|
| # ? May 18, 2013 19:41 |
|
|
Bob Morales posted:Does any of you guys read the Linux magazines? I'm hoping they make a lot of these available on Newsstand for iOS, along with a few other IT magazines. We'll see.
|
| # ? Oct 6, 2011 19:05 |
|
