[gentoo-user] [OT] Infrastructure documentation
Hello yall! This is half-topic but I think everyone will benefit from the results. Here at my company we have 18 Gentoo servers spread over 3 data centers and our office. We have well defined and time-proven (one of the servers was installed in 2004 and it's still the same Gentoo) processes on monitoring, backup, applying security fixes, and maintenance. We have almost 100% of high availability and some services even have high availability across different data centers in Florida and California. We are using catalyst with cfengine to save us a few hours of work every week and everything is working great. I would like to move forward to some other projects but most of the knowledge (60%) required to do everything resides on my head alone. I'm hit by a car in the streets and something might go bad, like the required monthly database partition maintenance. I would like to hear from the list what you are using for infrastructure, software, processes, hardware documentation. I think I need a system with a good user access control, an all-in-one solution to document everything. Using Wiki+UML would solve the issue (I have 30% already documented in wikis) but (a) none of them were designed for this specific task and (b) they don't integrate, people would have to use two systems that knows nothing about each other. What are you gurus doing to be replaceable? Thank you very much in advance! Best regards, Daniel Colchete
[gentoo-user] [OT] Infrastructure Documentation
Hello yall! This is half-topic but I think everyone will benefit from the results. Here at my company we have 18 Gentoo servers spread over 3 data centers and our office. We have well defined and time-proven (one of the servers was installed in 2004 and it's still the same Gentoo) processes on monitoring, backup, applying security fixes, and maintenance. We have almost 100% of high availability and some services even have high availability across different data centers in Florida and California. We are using catalyst with cfengine to save us a few hours of work every week and everything is working great. I would like to move forward to some other projects but most of the knowledge (60%) required to do everything resides on my head alone. I'm hit by a car in the streets and something might go bad, like the required monthly database partition maintenance. I would like to hear from the list what you are using for infrastructure, software, processes, hardware documentation. I think I need a system with a good user access control, an all-in-one solution to document everything. Using Wiki+UML would solve the issue (I have 30% already documented in wikis) but (a) none of them were designed for this specific task and (b) they don't integrate, people would have to use two systems that knows nothing about each other. What are you gurus doing to be replaceable? Thank you very much in advance! Best regards, Daniel Colchete PS: sending it again but in plain text this time. sorry :-) -- gentoo-user@lists.gentoo.org mailing list
Re: [gentoo-user] Linux too damn slow if memory 3GB
On 7/28/07, Stroller [EMAIL PROTECTED] wrote: On 29 Jul 2007, at 00:28, Daniel van Ham Colchete wrote: ... The fact is, if I boot with mem=3072M everything goes as fast as it should but I'm not using 1GB of memory. If I don't put the mem option, Linux will see 4GB of memory available but it will be damn slow (really). ... My processor is a Intel Core 2 Duo E6320 (1.86Ghz). The motherboard is Intel. I'm running Gentoo at 32bits mode and the kernel version is 2.6.20-gentoo-r8. Hi there, Does your Intel motherboard feature the 945PM chipset? I read about this last week: http://listserver.themacintoshguy.com/pipermail/x4u/2007-July/ 018031.html Stroller. -- [EMAIL PROTECTED] mailing list Stoller, actually it's the 965, but the link Tim sent me shows that it's not a chipset limitation... It's a 32bits design limitation... Best, Daniel
Re: [gentoo-user] Linux too damn slow if memory 3GB
On 7/30/07, Developer Edoceo [EMAIL PROTECTED] wrote: On 7/30/07, Steen Eugen Poulsen [EMAIL PROTECTED] wrote: Daniel van Ham Colchete skrev: actually it's the 965, but the link Tim sent me shows that it's not a chipset limitation... It's a 32bits design limitation... Thats not true, I run 32bit Gentoo with 4 GB Memory and has no slowdown issues I can measure with the naked eye. Seems to me like you have some hardware problem. I have two Gentoo systems (2.6.x) one with 4G and another with 8G. Both systems are are wicked fast. On the 4G `free` shows all but 21K of physical used, on the 8G there's still 2G free and the system is highly responsive. Can't remember which MB chipset I have, sorry. multiple Intel multicore cpu in both. What kernel version are you using?
[gentoo-user] Linux too damn slow if memory 3GB
Hello everyone!!! May be somebody can shed some light here... I'm building a server here with 4GB of RAM memory. The fact is, if I boot with mem=3072M everything goes as fast as it should but I'm not using 1GB of memory. If I don't put the mem option, Linux will see 4GB of memory available but it will be damn slow (really). To make Linux recognize 4 gigs of memory I had to activate HIGHMEM=64GB, otherwise it only recognizes 3279MB (but it is fast). My processor is a Intel Core 2 Duo E6320 (1.86Ghz). The motherboard is Intel. I'm running Gentoo at 32bits mode and the kernel version is 2.6.20-gentoo-r8. Any lights would be very appreciated. Best regards, Daniel Colchete
Re: [gentoo-user] Linux too damn slow if memory 3GB
On 7/28/07, Tim Allingham [EMAIL PROTECTED] wrote: On Sun, 2007-07-29 at 09:39 +1000, Tim Allingham wrote: I don't actually run 4GB of RAM in any of my intel systems, so I can't comment from experience, however my suspicion would be the overhead introduced from PAE, which (at least on older kernels) requires some processing overhead to utilise. Are you able to try a 64-bit install to determine if this is the case? Regards, Tim Allingham Email:[EMAIL PROTECTED] just to add, this page gives a reasonable explanation of the problems with 32-bit OS's trying to address large RAM volumes http://www.dansdata.com/askdan00015.htm That is very very very sad... The fact is I can't use 64bits for now... So let's take one gig out! Thanks for your help Daniel
[gentoo-user] Using SFQ for fair bandwidth alocation on servers
Hello yall! I have a collocation server with a 2 Mbps bandwidth for my e-mail. I would like to use SFQ to do a more fair split of this bandwidth between my users (~500). I think SFQ is a very good idea: it will split the available bandwidth between the current connections (almost) equally, and it's CPU cheap. The problem is: from what I could read, SFQ can only read the available bandwidth from the physical layer. Is it possible to tell SFQ something like I only have 2 Mbps, and not 100 Mbps, from eth0? Maybe if I put SFQ inside another queue that has everything? Best regards, Daniel Colchete
Re: [gentoo-user] Remote administration of a server
On 5/14/07, Hans-Werner Hilse [EMAIL PROTECTED] wrote: Hi, On Mon, 14 May 2007 15:42:45 -0300 Daniel van Ham Colchete [EMAIL PROTECTED] wrote: Thinking about other options, does anyone have any other tip for me? Am I going in the right direction? The two options you've mentioned are quite different. One gives console access, the other basically cures HD fails. The latter is clearly a job for your hosting company, I think. And there's an old, proven way for the task console access: forget about that graphics output on that computer and learn to trust in good ol' serial connections :-) certainly cheaper than KVM-over-IP. Another option would be for the servers to default to netbooting and fall back to HD on boot. Then you were able to switch on the service offering the netboot images on some fall-back servers on-demand. I think this is somewhat like your USB idea. Or generally use netboot (w/ redundant servers) and forget about the HD fails alltogether (i.e., have some remote login program in your initrd). All these options still won't give you the opportunity to power-cycle your machines, which might be the only option left under some circumstances. A hw watchdog can probably reduce the impact of that problem a lot. -hwh -- [EMAIL PROTECTED] mailing list Hi Hans! Yeah! Direct netboot is a very nice idea too... I can't to it with one kind of server I'll have but with the clustered ones that I'll be nice! To improve reliability I could make a copy of a healthy image to the node's hard drive every time it boots, so it's not dependent on a NFS server all the time (just to boot). Best Daniel
Re: [gentoo-user] Remote administration of a server
On 5/13/07, Mike Williams [EMAIL PROTECTED] wrote: On Monday 14 May 2007 19:42:45 Daniel van Ham Colchete wrote: To solve this, I'm seeing 2 options right now. The first would be buying a KVM-over-IP unit. But, a KVM-over-IP unit with the number of ports I need is expensive, almost as expensive as the servers it will be connected. We use a Belkin KVM-IP box with 2 sets of ports. One set for a local keyboard/monitor/mouse, and one to a 16 port Belkin KVM. The IP box can handle upto 64 KVM ports, and the KVM can daisy chain to an extent I don't know. You don't say how many servers, but 64 is quite a lot. I believe Belkin do a 4 channel IP box too. Total cost was, I think, ~£700. To remotely manage 16 servers, peanuts. That's cheap! I'm looking at their page right now. And 64 is enough for sure (at least during the first year).
[gentoo-user] Remote administration of a server
Hello yall, first, this is a kind of half-topic issue, but I'll be using Gentoo and the matter is on the interest of this list. I'll be building a Gentoo Cluster soon in a datacenter 6000 miles (9600 km) away from me... This project has to be as cost efficient as possible. A lot of research was made in this heading: the best cost effective solution. Everything I'll be redundant and scalable. Somethings have three levels of fail safeness (like my storage). So everything can fail. Every single item on the cluster can fail and my service will still be online. Right now I'm concerned with how I'm going to fix software problems when they arrive. I'm thinking about a situation where I have a kernel panic or when the Linux won't boot for any reason (incorrect kernel upgrade, hard drive failure, etc...). To solve this, I'm seeing 2 options right now. The first would be buying a KVM-over-IP unit. But, a KVM-over-IP unit with the number of ports I need is expensive, almost as expensive as the servers it will be connected. The second option would be having another server acting as a USB Guest. This usb-guest-server would be connected to every other server through a USB cable and would be seen as a pen drive with a Gentoo rescue disk inside. Them, if something goes wrong, I can activate the virtual pen drive, remotely reboot the troubled server and it will boot the pen drive. There is a howto about this at http://www.linux-usb.org/gadget/file_storage.html. But I can't find the necessary hardware to do this. Has anyone been able to do anything like this? Thinking about other options, does anyone have any other tip for me? Am I going in the right direction? For the obvious answer: I know it's better to be closer to the datacenter, but that's not an option for me right now and I know I'll have a remote-hands service, but it can be very time inefficient sometimes and I'm trying to avoid it as much as possible. Thank you all. Best regards, Daniel Colchete
[gentoo-user] SCIRE Project
Hello everyone Here on my company we are going to start deploying Gentoo Linux on our customers. Every server will have the very same installed packages, the very same use flags, very same cflags, only a few configurations will differ. I would like the deployment and maintenance to be done as easily as possible because this project needs to be scalable to more than 100 servers. Although we are going to install only 10 servers in the beginning, my boss says that I should be prepared for this number to grow. Yesterday I found about the SCIRE project that seems to solve my problems easily. But it seems that the project's development is stopped. Unfortunately, I don't know a thing of Phyton, so I can't help. Do anyone know how is the project going? Are we going to have a production usable release? If so, when? It's not like I'm pushing anything, I just want to know if I can count on it or not. Setting the project aside, I'm thinking about developing my own installer to install a catalyst's stage4 and reboot a working Gentoo. After that I'm thinking about using emerge with binary packages to install updates automatically. What do you think? Will it work? Is it possible to rollback an update if something goes wrong? To solve the problem with incompatible configuration files, everytime I upgrade anything, a perl script will reconfigure the customers server. Thank you all for your help and insights. Best regards, Daniel Colchete -- gentoo-user@gentoo.org mailing list
Re: [gentoo-user] Gigabit NIC but only 100baseT/Full working
On 12/4/06, kashani [EMAIL PROTECTED] wrote: Daniel van Ham Colchete wrote: Hi all! I have two servers here with two Intel E1000 NIC each. I'm planing on using the eth1 ethernet with DRBD. My problem is that although the NICs support gigabit ethernet, they only negociate with 100baseT/Full. I'm using a CAT-5E crossover cable between the servers. Does anyone have any tip? auto negotiation sometimes doesn't though it's light years better than it was ten years ago. I'd play around with ethool -s and set both sides manually to full 1000. If you don't get errors I'd make that part of the boot process as well as the network scripts if you restart your interfaces often. I tried that and didn't work. But it's working now, see my next msg. Daniel -- gentoo-user@gentoo.org mailing list
Re: [gentoo-user] Gigabit NIC but only 100baseT/Full working
On 12/5/06, Pawel Kraszewski [EMAIL PROTECTED] wrote: Dnia poniedziałek, 4 grudnia 2006 22:49, Daniel van Ham Colchete napisał: I'm planing on using the eth1 ethernet with DRBD. My problem is that although the NICs support gigabit ethernet, they only negociate with 100baseT/Full. I'm using a CAT-5E crossover cable between the servers. I'm not 100% sure, but try a straight cable. I don't quite remeber, but this is due to that all 4 pairs are bi-directional and/or 1k cards are MDI/MDIX autosensing. OTOH see http://www.ertyu.org/steven_nikkel/ethernetcables.html Pawel, you were right. A straight cable worked perfectly! :-)! Now my network autonegotiated at 1000baseT/Full. Good reference too. Thanks! Best, Daniel -- gentoo-user@gentoo.org mailing list
[gentoo-user] Gigabit NIC but only 100baseT/Full working
Hi all! I have two servers here with two Intel E1000 NIC each. I'm planing on using the eth1 ethernet with DRBD. My problem is that although the NICs support gigabit ethernet, they only negociate with 100baseT/Full. I'm using a CAT-5E crossover cable between the servers. Here is my dmesg (both servers): e1000: eth1: e1000_probe: Intel(R) PRO/1000 Network Connection e1000: eth1: e1000_watchdog: NIC Link is Up 100 Mbps Full Duplex My ethtool shows the following (both servers): Settings for eth1: Supported ports: [ TP ] Supported link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full Supports auto-negotiation: Yes Advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full Advertised auto-negotiation: Yes Speed: 100Mb/s Duplex: Full Port: Twisted Pair PHYAD: 0 Transceiver: internal Auto-negotiation: on Supports Wake-on: umbg Wake-on: g Current message level: 0x0007 (7) Link detected: yes Does anyone have any tip? Best regards, Daniel Colchete -- gentoo-user@gentoo.org mailing list
Re: [gentoo-user] Gigabit NIC but only 100baseT/Full working
Thanks Rick, I'm going to try that tomorrow. I thought CAT5e cable would support gigabit connections. Best regards, Daniel Colchete On 12/4/06, Khabi [EMAIL PROTECTED] wrote: From my understanding thats not a problem with the OS, but with the crossover cable itself. The os is seeing it as a Gigabit card. I believe a cat6 cable will work and give you the right speed, you may want to try that. Alternativly, the following should work if you crimp one yourself. Connector1 Connector2 -- Orange/whiteGreen/White Orange Green Green/White Orange/White BlueBrown/White Blue/White Brown Green Orange Brown/White Blue Brown Blue/White -Rick -- gentoo-user@gentoo.org mailing list