RE: [asterisk-users] zaptel huge irq problem
I'm no Asterisk expert (nor Linux expert) but I am using my * box for multiple things (transparent firewall, NAT box, samba server, poptop server) and for a considerable time I've been running a VmWare server with a Windows XP virtual machine up-and-running at all times! The Windows XP VM was running IIS, Apache, WarFTP and a Firebird database server - all of which got moderate use. The hardware for my * box is what would be considered moderate-to-cheap: Sempron-something processor (not a big processor, don't remember the exact GHz), enough RAM (I've added 1 Gb of RAM when I've started using VmWare server), a nice motherboard (I remember I specifically looked for a motherboard with the minimum amount of on-board devices, of which I have disabled everything I don't need!). The extra hardware on my box includes 2 PCI NIC's (I'm also using the on-board NIC so I've got 3 working NIC's), an TDM400 card with 3 FXO and 1 FXS, and an Diva Eicon Server BRI card for my ISDN connection. I've got 3 HDD's into the box, of which 2 are old IDE drivers (parallel ATA) and the other one is SATA. My VoIP experience has been good, my zaptel timing is pretty good and I can get faxes working on the FXS interface as well (coming in over the ISDN line). The rationale behind placing the Windows XP virtual machine on the * has not been the lack of extra hardware but the desire to keep the number of always-on servers to a minimum. I've since moved the VM off the Asterisk server because I've installed an Windows SBS 2003 server on a considerably more powerful server. So there it goes, proof that a small-office Asterisk box can do lots and lots of things! -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of François Delawarde Sent: Thursday, May 17, 2007 8:26 PM To: Asterisk Users Mailing List - Non-Commercial Discussion Subject: Re: [asterisk-users] zaptel huge irq problem Hi, Why are you so determined to use Asterisk in a VM? You're asking for trouble. Asterisk belongs on dedicated hardware. I actually want to use Asterisk in a machine HOSTING a VM (that's what I implied with the Dom-0 thing I said earlier), sorry for the misunderstanding. I agree with you that given the state of advancement of just about any 'virtualizer', I would have to be totally stupid to try running Asterisk inside a VM. (I also wouldn't have asked here in the first place, as I would have been totally certain that problems came from the virtualizer itself) If you feel concerned with my reasons for doing that anyway: - No one told me that Asterisk belonged on dedicated hardware before you, so I didn't know. - I'm just not very rich and try to integrate some things I need in my machine (don't worry, I did not framebuffered or X.orged it yet) because I cannot afford to buy another one (yes, even the 200€ one)... The part you don't want to know is how many people I had to kill in order to get my TDM400 card, until I found out that other cheaper solutions existed. :-) We're just trying to help -- but if you insist on running Asterisk in a VM, then you're on your own. And I thank you for that (the helping part), you've found the deep cause of all my zaptel problems (Xen), so please don't leave me alone! ;-) To be a bit more constructive, I'd like to ask you or anyone that dared to try using Asterisk on a non-dedicated hardware, specifically those that tried on a machine hosting VMs the following: - If there is no way running Asterisk with Xen, what type of 'hypervisor' should I use in order not to have problems? KVM?, KQemu?, VMWare? - What type of problems should I expect if I dare to do that? (of course, Asterisk will be realtime-niced to make it more important) Thanks and sorry again for the misunderstandings, François. ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
Hi, I don't want multiple instances of Asterisk. My goal here is to make Asterisk and its Zaptel hardware run nice on a machine that is not dedicated and also hosts VM. I had lot's of problems with Xen, as the host runs a modified kernel that has apparently issues with the interrupt handling (at least). My question was more of what kind of hypervisor I should use for Asterisk/Zaptel not to have problems like it has in Xen. François. Tzafrir Cohen wrote: On Thu, May 17, 2007 at 07:26:10PM +0200, François Delawarde wrote: Hi, Why are you so determined to use Asterisk in a VM? You're asking for trouble. Asterisk belongs on dedicated hardware. I actually want to use Asterisk in a machine HOSTING a VM (that's what I implied with the Dom-0 thing I said earlier), sorry for the misunderstanding. I agree with you that given the state of advancement of just about any 'virtualizer', I would have to be totally stupid to try running Asterisk inside a VM. (I also wouldn't have asked here in the first place, as I would have been totally certain that problems came from the virtualizer itself) What kind of separation do you really need? Xen, VMWare and such are big cannons here. Every virtual machine will consume fixed ammount of memory. There is a considerable overhead for hardware access. It allows you things like running different OS/distribution on each guest. But for some reason I'm not sure you really need that? Will the users have direct acces to the dialplan and the rest of the configuration? If not: just run a single instance of Asterisk. If you do need multiple asterisk instances, verver or openvz might help you to give a separate container for that user's personal usage. Stephan has mentioned in this thread he set up several Asterisk-es on a vserver system. -- _ François Delawarde Ingeniero de red Tel: 918.03.92.51 E-mail: [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] _ WIRELESS MUNDI http://www.wirelessmundi.com/ C/Isaac Newton, 1 - Oficina 26 · Parque Tecnológico de Madrid 28760 TRES CANTOS (Madrid) Tlf./Fax: (+34) 918 03 92 51 La información contenida en este mensaje y en sus archivos adjuntos es CONFIDENCIAL y se dirige exclusivamente a sus destinatarios. Queda expresamente prohibida la utilización de la misma por cualquier persona distinta de los destinatarios de esta comunicación. Si usted ha recibido este mensaje por error le rogamos que lo comunique inmediatamente a WIRELESS MUNDI y lo borre al igual que todos sus documentos adjuntos. El correo electrónico no puede asegurar la confidencialidad ni la integridad de sus mensajes por lo que WIRELESS MUNDI no se hace responsable de tales errores u omisiones. --0-- All information in this message and its attachments is confidential and may be legally privileged. Only intended recipients are authorized to use it. If you have received this transmission in error, please notify WIRELESS MUNDI immediately and delete this message and its attachments. E-mail transmissions are not guaranteed to be secure or error free and WIRELESS MUNDI does not accept liability for such errors or omissions. ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
I will try VMWare then. Does it support AMD-V or Intel VT virtualization at processor level for speedup? If so, I guess there is at some point some kernel code (module like KVM or special kernel like Xen) that could provoke some similar problems with zaptel interrupts. Could there be some issues here or am I totally wrong again? François. Stephen Bosch wrote: François Delawarde wrote: And I thank you for that (the helping part), you've found the deep cause of all my zaptel problems (Xen), so please don't leave me alone! ;-) To be a bit more constructive, I'd like to ask you or anyone that dared to try using Asterisk on a non-dedicated hardware, specifically those that tried on a machine hosting VMs the following: - If there is no way running Asterisk with Xen, what type of 'hypervisor' should I use in order not to have problems? KVM?, KQemu?, VMWare? The only one I would bother with is VMWare Server. It is solid, proven technology, and they have a big team of very talented engineers who have worked years to get the virtualization to the point where it can be sold as an enterprise grade product. If I were to try virtualizing anything, it would be on VMWare Server. - What type of problems should I expect if I dare to do that? (of course, Asterisk will be realtime-niced to make it more important) Well, in particular anything that expects unfettered access to hardware (as most realtime applications which rely on interface cards do) is going to be vulnerable to the proclivities of the hypervisor. Virtualization is still mostly rocket science. I have no doubt that it is the future and one day everything will run in virtualized environments -- but we're still a bit away from that. Virtualization makes financial sense when you have 20 database servers running at 10% utilization; you can drop your hardware requirements by at least a third... but for systems relying on dedicated hardware, I would be very careful (again -- I speak from ugly experience here). -Stephen- ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users -- _ François Delawarde Ingeniero de red Tel: 918.03.92.51 E-mail: [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] _ WIRELESS MUNDI http://www.wirelessmundi.com/ C/Isaac Newton, 1 - Oficina 26 · Parque Tecnológico de Madrid 28760 TRES CANTOS (Madrid) Tlf./Fax: (+34) 918 03 92 51 La información contenida en este mensaje y en sus archivos adjuntos es CONFIDENCIAL y se dirige exclusivamente a sus destinatarios. Queda expresamente prohibida la utilización de la misma por cualquier persona distinta de los destinatarios de esta comunicación. Si usted ha recibido este mensaje por error le rogamos que lo comunique inmediatamente a WIRELESS MUNDI y lo borre al igual que todos sus documentos adjuntos. El correo electrónico no puede asegurar la confidencialidad ni la integridad de sus mensajes por lo que WIRELESS MUNDI no se hace responsable de tales errores u omisiones. --0-- All information in this message and its attachments is confidential and may be legally privileged. Only intended recipients are authorized to use it. If you have received this transmission in error, please notify WIRELESS MUNDI immediately and delete this message and its attachments. E-mail transmissions are not guaranteed to be secure or error free and WIRELESS MUNDI does not accept liability for such errors or omissions. ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
I don't really know of other virtualization technology other than Xen, and I thank you for guiding me through this, but I have a few doubts related to the choice of a virtualization technology in a host with Asterisk: - Isn't the fact that KVM is now included in the mainstream Linux kernel as of 2.6.20 a certain type of 'proof' that it could be stable enough compared to others (of course there could be licensing or other political/friendship issues)? - Even if the virtual guests aren't totally stable and 100% reliable yet, wouldn't the use of KVM be better with Zaptel compatible cards than Xen, in architecture point of vue, as it is only a kernel module that -as far as I know- don't appear to be changing fundamental issues like IRQ handling or I/O scheduling in the kernel, and from the fact that virtual machines are treated like simple processes? François. Andrew Kohlsmith wrote: On Wednesday 16 May 2007 1:00 pm, François Delawarde wrote: Thanks again for your help, and sorry if I was not 'that' convinced on your first answer and sent a mail to Xen user mailing list to check if they knew that issue (no answer yet). Now I almost believe you a lot. If I understand well I have two options, recode Xen or abandon it. I'll probably go for the 2nd choice and start looking at other solutions, KVM seems to be a good choice and shouldn't interfere much with Asterisk (again: as far as mailing lists say). Let me try to understand this: Xen is a (far) more mature virtualization technology than KVM, and it's been said that it's commercialization was rushed. So you're going to try KVM, which is still under heavy development, as a stable solution? -A. ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
François Delawarde wrote: I don't really know of other virtualization technology other than Xen, and I thank you for guiding me through this, but I have a few doubts related to the choice of a virtualization technology in a host with Asterisk: - Isn't the fact that KVM is now included in the mainstream Linux kernel as of 2.6.20 a certain type of 'proof' that it could be stable enough compared to others (of course there could be licensing or other political/friendship issues)? It is anything but proof. Kernel inclusions are often a matter of convenience. - Even if the virtual guests aren't totally stable and 100% reliable yet, wouldn't the use of KVM be better with Zaptel compatible cards than Xen, in architecture point of vue, as it is only a kernel module that -as far as I know- don't appear to be changing fundamental issues like IRQ handling or I/O scheduling in the kernel, and from the fact that virtual machines are treated like simple processes? Why are you so determined to use Asterisk in a VM? You're asking for trouble. Asterisk belongs on dedicated hardware. We're just trying to help -- but if you insist on running Asterisk in a VM, then you're on your own. That's not a risk I'd want to take. -Stephen- ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
Hi, Why are you so determined to use Asterisk in a VM? You're asking for trouble. Asterisk belongs on dedicated hardware. I actually want to use Asterisk in a machine HOSTING a VM (that's what I implied with the Dom-0 thing I said earlier), sorry for the misunderstanding. I agree with you that given the state of advancement of just about any 'virtualizer', I would have to be totally stupid to try running Asterisk inside a VM. (I also wouldn't have asked here in the first place, as I would have been totally certain that problems came from the virtualizer itself) If you feel concerned with my reasons for doing that anyway: - No one told me that Asterisk belonged on dedicated hardware before you, so I didn't know. - I'm just not very rich and try to integrate some things I need in my machine (don't worry, I did not framebuffered or X.orged it yet) because I cannot afford to buy another one (yes, even the 200€ one)... The part you don't want to know is how many people I had to kill in order to get my TDM400 card, until I found out that other cheaper solutions existed. :-) We're just trying to help -- but if you insist on running Asterisk in a VM, then you're on your own. And I thank you for that (the helping part), you've found the deep cause of all my zaptel problems (Xen), so please don't leave me alone! ;-) To be a bit more constructive, I'd like to ask you or anyone that dared to try using Asterisk on a non-dedicated hardware, specifically those that tried on a machine hosting VMs the following: - If there is no way running Asterisk with Xen, what type of 'hypervisor' should I use in order not to have problems? KVM?, KQemu?, VMWare? - What type of problems should I expect if I dare to do that? (of course, Asterisk will be realtime-niced to make it more important) Thanks and sorry again for the misunderstandings, François. ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
Hi, I want to quickly mention that I've had great success with running Asterisk in the under-appreciated Linux-VServer environment. This is not so much a virtualisation environment as a system partioner on steroids. Nothing to do with running windows on Linux and suchlike, but a good way to run lots of Asterisk and other stuff isolated from each other. There is only one kernel, and hardware is not virtualised. A particular guest We run about 10 Asterisk instances, together with web servers, Mysql and more. All on a 1GB RAM Pentium D box. CPU and memory to spare. hildegard steve # vserver-stat CTX PROCVSZRSS userTIME sysTIMEUPTIME NAME 0 82 266.5M 20.6M 21h19m47 8h04m51 26d02h24 root server 8 63 710.6M 1.4G 14m10s98 5m31s37 26d02h20 ctel_web 9 13 1.1G 31.8M 16m28s43 14m02s46 25d05h57 ctel_pbx 10 19 694.8M 172.7M 2h01m43 19m59s65 20d07h58 voipconnect 11 8 927.9M 98.9M 40m43s42 10m09s30 26d02h21 ctel_admin 12 5 210.5M 21.5M 5h13m35 36m41s61 26d02h21 ctel_db 13 8 903.4M 55.2M 3m00s10 1m09s40 26d02h20 ctel_intranet 15 5 213M 1.2M 9m09s35 12m51s00 26d02h19 xconnect 33 29 261.2M 22.7M 0m07s65 0m09s13 3d08h43 testtrunk 56 13 1.1G10M 1m44s45 1m18s94 26d02h21 aaa 57 13 1.1G 13.9M 8m47s52 8m46s10 26d02h21 bbb 58 13 1.1G 16.1M 54m29s99 24m46s22 26d02h14 ccc 60 9 293.4M 31.9M 10h44m42 2h00m55 26d02h20 ddd 61 13 1.1G 6.7M 12m45s11 13m26s42 26d02h19 eee (26 days uptime? Our hosting provider had to do power maintenance. We've never had a crash of the host system). My zttest inside a guest machine: voipconnect zaptel # ./zttest Opened pseudo zap interface, measuring accuracy... 100.00% 99.987793% 100.00% 99.987793% 100.00% 100.00% 99.987793% 100.00% 100.00% 99.987793% 100.00% 99.987793% 100.00% 100.00% 99.987793% 100.00% 99.987793% 100.00% 99.987793% 100.00% 100.00% 99.987793% 100.00% 99.987793% 100.00% 100.00% 99.987793% 100.00% 99.987793% 100.00% 100.00% --- Results after 31 passes --- Best: 100.00 -- Worst: 99.987793 -- Average: 99.995277 Try that with Xen or VMWare. http://www.linux-vserver.org/ (Our host is hardened gentoo with PaX and GRSecurity, plus vserver; guests are gentoo too, though VServer does support guests being different distributions). Hope this helps someone, Steve Davies ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
On 14/05/07, Salvatore Giudice [EMAIL PROTECTED] wrote: Try switching to a Sangoma card. You won't have anymore IRQ issues once you abandon Digium hardware. Its not true, by the way. I've assisted more than one person using a Sangoma who was having issues caused by interrupt stuff. And it was the same sort of things that might affect a Digium board- motherboard raid disabling interrupts, sharing an IRQ with a heavy-interrupting LAN card, etc. Not suprising since its the same underlying problem - excessing interrupt handling latency. Steve ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
François Delawarde wrote: And I thank you for that (the helping part), you've found the deep cause of all my zaptel problems (Xen), so please don't leave me alone! ;-) To be a bit more constructive, I'd like to ask you or anyone that dared to try using Asterisk on a non-dedicated hardware, specifically those that tried on a machine hosting VMs the following: - If there is no way running Asterisk with Xen, what type of 'hypervisor' should I use in order not to have problems? KVM?, KQemu?, VMWare? The only one I would bother with is VMWare Server. It is solid, proven technology, and they have a big team of very talented engineers who have worked years to get the virtualization to the point where it can be sold as an enterprise grade product. If I were to try virtualizing anything, it would be on VMWare Server. - What type of problems should I expect if I dare to do that? (of course, Asterisk will be realtime-niced to make it more important) Well, in particular anything that expects unfettered access to hardware (as most realtime applications which rely on interface cards do) is going to be vulnerable to the proclivities of the hypervisor. Virtualization is still mostly rocket science. I have no doubt that it is the future and one day everything will run in virtualized environments -- but we're still a bit away from that. Virtualization makes financial sense when you have 20 database servers running at 10% utilization; you can drop your hardware requirements by at least a third... but for systems relying on dedicated hardware, I would be very careful (again -- I speak from ugly experience here). -Stephen- ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
On Thu, May 17, 2007 at 07:26:10PM +0200, François Delawarde wrote: Hi, Why are you so determined to use Asterisk in a VM? You're asking for trouble. Asterisk belongs on dedicated hardware. I actually want to use Asterisk in a machine HOSTING a VM (that's what I implied with the Dom-0 thing I said earlier), sorry for the misunderstanding. I agree with you that given the state of advancement of just about any 'virtualizer', I would have to be totally stupid to try running Asterisk inside a VM. (I also wouldn't have asked here in the first place, as I would have been totally certain that problems came from the virtualizer itself) What kind of separation do you really need? Xen, VMWare and such are big cannons here. Every virtual machine will consume fixed ammount of memory. There is a considerable overhead for hardware access. It allows you things like running different OS/distribution on each guest. But for some reason I'm not sure you really need that? Will the users have direct acces to the dialplan and the rest of the configuration? If not: just run a single instance of Asterisk. If you do need multiple asterisk instances, verver or openvz might help you to give a separate container for that user's personal usage. Stephan has mentioned in this thread he set up several Asterisk-es on a vserver system. -- Tzafrir Cohen icq#16849755jabber:[EMAIL PROTECTED] +972-50-7952406 mailto:[EMAIL PROTECTED] http://www.xorcom.com iax:[EMAIL PROTECTED]/tzafrir ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
aaah... I'm running asterisk in a Xen kernel, but not on a virtual machine (DomU), only on Dom0, so it's supposed to be running on the physical server (no PCI frontend device, ...). I had seen possible problems with older versions of Xen, but only with ztdummy timing and on DomU virtual machines. On Xen mailing list, they all advise to use a digium PCI card to remove those problems. I will test and report what happens with a normal kernel, but meanwhile doesn't anyone know of a possible possibility to make it work with this setting playing for example with IRQ priorities or something, or isn't there any hope at all? Thanks again, François. Stephen Bosch wrote: François Delawarde wrote: Thanks Michael, I've already been through all that unfortunately, and I have a SATA drive, so no UDMA mode 2 as far as I know. I'm currently trying everything again anyway, but i doubt it will work if nothing worked the first time. Anyone would know of issues with XEN or SMP (or both) kernel? Do dual core AMD64 processors have issues? Uh, yeah... Xen has many, many problems with interrupt handling and is utterly unsuitable for running anything that depends on hardware peripherals. I speak from very painful experience. There is no way, under any circumstance, that I would try to run Asterisk with interface cards in a Xen environment. It's too bad you wasted so much time trying to fix it -- it's never going to work. Try ripping Xen out and doing it directly on the physical server. I think you'll find your problems will go away. -Stephen- ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
François Delawarde wrote: aaah... I'm running asterisk in a Xen kernel, but not on a virtual machine (DomU), only on Dom0, so it's supposed to be running on the physical server (no PCI frontend device, ...). I had seen possible problems with older versions of Xen, but only with ztdummy timing and on DomU virtual machines. On Xen mailing list, they all advise to use a digium PCI card to remove those problems. You had seen? Did you have these problems personally, or are you going by mailing list postings? I will test and report what happens with a normal kernel, but meanwhile doesn't anyone know of a possible possibility to make it work with this setting playing for example with IRQ priorities or something, or isn't there any hope at all? We abandoned Xen (recent versions too!) after serious interrupt problems (it doesn't matter if you are in domU or dom0, by the way) that caused the entire *system*, with all the VMs, to lock up *hard* whenever we started to push significant amounts of data through anywhere, be it an Ethernet controller or a SCSI adapter. It is in need of a lot of work. Their efforts to commercialize it are premature. We had to learn this hard way, unfortunately. If you need virtualization that badly, you might want to consider going with VMware Server, which is now freely available. My experience with VMware has been better. -Stephen- ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
Stephen Bosch wrote: François Delawarde wrote: aaah... I'm running asterisk in a Xen kernel, but not on a virtual machine (DomU), only on Dom0, so it's supposed to be running on the physical server (no PCI frontend device, ...). I had seen possible problems with older versions of Xen, but only with ztdummy timing and on DomU virtual machines. On Xen mailing list, they all advise to use a digium PCI card to remove those problems. You had seen? Did you have these problems personally, or are you going by mailing list postings? Sorry for not being Scottish anymore, my English is not what it used to be a few hundred years ago. I meant that I'm mainly going by mailing list postings. I will test and report what happens with a normal kernel, but meanwhile doesn't anyone know of a possible possibility to make it work with this setting playing for example with IRQ priorities or something, or isn't there any hope at all? We abandoned Xen (recent versions too!) after serious interrupt problems (it doesn't matter if you are in domU or dom0, by the way) that caused the entire *system*, with all the VMs, to lock up *hard* whenever we started to push significant amounts of data through anywhere, be it an Ethernet controller or a SCSI adapter. It is in need of a lot of work. Their efforts to commercialize it are premature. We had to learn this hard way, unfortunately. If you need virtualization that badly, you might want to consider going with VMware Server, which is now freely available. My experience with VMware has been better. Thanks again for your help, and sorry if I was not 'that' convinced on your first answer and sent a mail to Xen user mailing list to check if they knew that issue (no answer yet). Now I almost believe you a lot. If I understand well I have two options, recode Xen or abandon it. I'll probably go for the 2nd choice and start looking at other solutions, KVM seems to be a good choice and shouldn't interfere much with Asterisk (again: as far as mailing lists say). François. ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
On Wednesday 16 May 2007 1:00 pm, François Delawarde wrote: Thanks again for your help, and sorry if I was not 'that' convinced on your first answer and sent a mail to Xen user mailing list to check if they knew that issue (no answer yet). Now I almost believe you a lot. If I understand well I have two options, recode Xen or abandon it. I'll probably go for the 2nd choice and start looking at other solutions, KVM seems to be a good choice and shouldn't interfere much with Asterisk (again: as far as mailing lists say). Let me try to understand this: Xen is a (far) more mature virtualization technology than KVM, and it's been said that it's commercialization was rushed. So you're going to try KVM, which is still under heavy development, as a stable solution? -A. ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
François Delawarde wrote: Stephen Bosch wrote: François Delawarde wrote: aaah... I'm running asterisk in a Xen kernel, but not on a virtual machine (DomU), only on Dom0, so it's supposed to be running on the physical server (no PCI frontend device, ...). I had seen possible problems with older versions of Xen, but only with ztdummy timing and on DomU virtual machines. On Xen mailing list, they all advise to use a digium PCI card to remove those problems. You had seen? Did you have these problems personally, or are you going by mailing list postings? Sorry for not being Scottish anymore, my English is not what it used to be a few hundred years ago. I meant that I'm mainly going by mailing list postings. No, there was nothing wrong with the grammar -- I was just suggesting that you not take the Xen mailing list postings too seriously. Better get it from someone with personal experience (and it sounds like you've had plenty of your own personal experience with this issue already). I don't know of anybody who is using it in a serious production environment anymore, for the reasons I've already noted. You'll notice that the release schedule has been, *ahem*, very slow; the mailing lists are littered with pleading posts from users reporting crashes and freeze-ups, sometimes catastrophic. Just have a look at the archives. (The Xensource people are also very quiet, which suggests to me that they themselves don't know how to address some of problems.) For hardware, it's just not good enough. It was an interesting academic project once, but like many such projects, its transition into the applied world has been rocky at best. I would love to be proven wrong, preferably sooner than later :) Thanks again for your help, and sorry if I was not 'that' convinced on your first answer and sent a mail to Xen user mailing list to check if they knew that issue (no answer yet). Now I almost believe you a lot. If I understand well I have two options, recode Xen or abandon it. I'll probably go for the 2nd choice and start looking at other solutions, KVM seems to be a good choice and shouldn't interfere much with Asterisk (again: as far as mailing lists say). Hey -- it's no skin off my nose if you want to keep using Xen. I just think you'll be wasting a lot of your time, which I'm sure is valuable. Cheers, -Stephen- ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
[asterisk-users] zaptel huge irq problem
Hello, I had noticed strange crackling sound on my phone calls going through my zaptel device (TDM400P), so i decided to check on possible timer issue, and found lots of issues on forums concerning the sensibility of zaptel with IRQs, and tried about everything: moving PCI slots, noapic and acpi=off boot options, play with different kernel options: iosched/preemption/timer/..., play with BIOS PCI options, change priorities, PCI latencies, IRQ balance, smp_afinity, but impossible to come up with anything correcting that problem. Any idea about this? Is it possible to force the timer to ztdummy (RTC timer) when you have a zap card plugged in? It's the only thing i could try to make it work. Thanks, François. Just in case: - Linux 2.6.18 with debian patches and xen enabled, asterisk running on dom0. - Here is my zttest results under a bit of load: # ./zttest Opened pseudo zap interface, measuring accuracy... 99.609375% 99.609375% 99.218750% 99.316406% 99.804688% 99.414062% 99.121094% 99.511719% 99.121094% 99.316406% 99.707031% 99.707031% 98.730469% 99.414062% 99.902344% 99.218750% 100.00% 99.414062% 98.828125% 99.218750% 99.316406% 98.449707% 100.00% - The card DOES NOT seem to share interrupts (checked also with lspci): # cat /proc/interrupts CPU0 CPU1 1: 1626 0Phys-irq i8042 6: 3 0Phys-irq floppy 8: 0 0Phys-irq rtc 9: 0 0Phys-irq acpi 14: 63 0Phys-irq ide0 16: 1 0Phys-irq libata, eth3 17:6762583 0Phys-irq libata 18: 13789 0Phys-irq libata 19: 33459690 0Phys-irq eth1 20: 19864325 0Phys-irq sky2, eth0 21: 269250881 0Phys-irq wctdm 256: 77735119 0 Dynamic-irq timer0 257:3986325 0 Dynamic-irq resched0 258: 37 0 Dynamic-irq callfunc0 259: 04652748 Dynamic-irq resched1 260: 0139 Dynamic-irq callfunc1 261: 0 28924306 Dynamic-irq timer1 262: 1021 0 Dynamic-irq xenbus 263: 0 0 Dynamic-irq console NMI: 0 0 LOC: 0 0 ERR: 0 MIS: 0 ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
RE: [asterisk-users] zaptel huge irq problem
François, I too had a similar problem and found the information on this page helpful: http://www.voip-info.org/wiki/view/Asterisk+PCI+bus+Troubleshooting What ended up working for me was changing the UDMA to mode 2 for the hard drive. Once I did that, this card has worked perfectly for me. Michael L. Young -Original Message- From: [EMAIL PROTECTED] [mailto:asterisk-users- [EMAIL PROTECTED] On Behalf Of François Delawarde Sent: Monday, May 14, 2007 10:24 AM To: asterisk-users@lists.digium.com Subject: [asterisk-users] zaptel huge irq problem Hello, I had noticed strange crackling sound on my phone calls going through my zaptel device (TDM400P), so i decided to check on possible timer issue, and found lots of issues on forums concerning the sensibility of zaptel with IRQs, and tried about everything: moving PCI slots, noapic and acpi=off boot options, play with different kernel options: iosched/preemption/timer/..., play with BIOS PCI options, change priorities, PCI latencies, IRQ balance, smp_afinity, but impossible to come up with anything correcting that problem. Any idea about this? Is it possible to force the timer to ztdummy (RTC timer) when you have a zap card plugged in? It's the only thing i could try to make it work. Thanks, François. Just in case: - Linux 2.6.18 with debian patches and xen enabled, asterisk running on dom0. - Here is my zttest results under a bit of load: # ./zttest Opened pseudo zap interface, measuring accuracy... 99.609375% 99.609375% 99.218750% 99.316406% 99.804688% 99.414062% 99.121094% 99.511719% 99.121094% 99.316406% 99.707031% 99.707031% 98.730469% 99.414062% 99.902344% 99.218750% 100.00% 99.414062% 98.828125% 99.218750% 99.316406% 98.449707% 100.00% - The card DOES NOT seem to share interrupts (checked also with lspci): # cat /proc/interrupts CPU0 CPU1 1: 1626 0Phys-irq i8042 6: 3 0Phys-irq floppy 8: 0 0Phys-irq rtc 9: 0 0Phys-irq acpi 14: 63 0Phys-irq ide0 16: 1 0Phys-irq libata, eth3 17:6762583 0Phys-irq libata 18: 13789 0Phys-irq libata 19: 33459690 0Phys-irq eth1 20: 19864325 0Phys-irq sky2, eth0 21: 269250881 0Phys-irq wctdm 256: 77735119 0 Dynamic-irq timer0 257:3986325 0 Dynamic-irq resched0 258: 37 0 Dynamic-irq callfunc0 259: 04652748 Dynamic-irq resched1 260: 0139 Dynamic-irq callfunc1 261: 0 28924306 Dynamic-irq timer1 262: 1021 0 Dynamic-irq xenbus 263: 0 0 Dynamic-irq console NMI: 0 0 LOC: 0 0 ERR: 0 MIS: 0 ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
Thanks Michael, I've already been through all that unfortunately, and I have a SATA drive, so no UDMA mode 2 as far as I know. I'm currently trying everything again anyway, but i doubt it will work if nothing worked the first time. Anyone would know of issues with XEN or SMP (or both) kernel? Do dual core AMD64 processors have issues? François. Michael L. Young wrote: François, I too had a similar problem and found the information on this page helpful: http://www.voip-info.org/wiki/view/Asterisk+PCI+bus+Troubleshooting What ended up working for me was changing the UDMA to mode 2 for the hard drive. Once I did that, this card has worked perfectly for me. Michael L. Young -Original Message- From: [EMAIL PROTECTED] [mailto:asterisk-users- [EMAIL PROTECTED] On Behalf Of François Delawarde Sent: Monday, May 14, 2007 10:24 AM To: asterisk-users@lists.digium.com Subject: [asterisk-users] zaptel huge irq problem Hello, I had noticed strange crackling sound on my phone calls going through my zaptel device (TDM400P), so i decided to check on possible timer issue, and found lots of issues on forums concerning the sensibility of zaptel with IRQs, and tried about everything: moving PCI slots, noapic and acpi=off boot options, play with different kernel options: iosched/preemption/timer/..., play with BIOS PCI options, change priorities, PCI latencies, IRQ balance, smp_afinity, but impossible to come up with anything correcting that problem. Any idea about this? Is it possible to force the timer to ztdummy (RTC timer) when you have a zap card plugged in? It's the only thing i could try to make it work. Thanks, François. Just in case: - Linux 2.6.18 with debian patches and xen enabled, asterisk running on dom0. - Here is my zttest results under a bit of load: # ./zttest Opened pseudo zap interface, measuring accuracy... 99.609375% 99.609375% 99.218750% 99.316406% 99.804688% 99.414062% 99.121094% 99.511719% 99.121094% 99.316406% 99.707031% 99.707031% 98.730469% 99.414062% 99.902344% 99.218750% 100.00% 99.414062% 98.828125% 99.218750% 99.316406% 98.449707% 100.00% - The card DOES NOT seem to share interrupts (checked also with lspci): # cat /proc/interrupts CPU0 CPU1 1: 1626 0Phys-irq i8042 6: 3 0Phys-irq floppy 8: 0 0Phys-irq rtc 9: 0 0Phys-irq acpi 14: 63 0Phys-irq ide0 16: 1 0Phys-irq libata, eth3 17:6762583 0Phys-irq libata 18: 13789 0Phys-irq libata 19: 33459690 0Phys-irq eth1 20: 19864325 0Phys-irq sky2, eth0 21: 269250881 0Phys-irq wctdm 256: 77735119 0 Dynamic-irq timer0 257:3986325 0 Dynamic-irq resched0 258: 37 0 Dynamic-irq callfunc0 259: 04652748 Dynamic-irq resched1 260: 0139 Dynamic-irq callfunc1 261: 0 28924306 Dynamic-irq timer1 262: 1021 0 Dynamic-irq xenbus 263: 0 0 Dynamic-irq console NMI: 0 0 LOC: 0 0 ERR: 0 MIS: 0 ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users -- _ François Delawarde Ingeniero de red Tel: 918.03.92.51 E-mail: [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] _ WIRELESS MUNDI http://www.wirelessmundi.com/ C/Isaac Newton, 1 - Oficina 26 · Parque Tecnológico de Madrid 28760 TRES CANTOS (Madrid) Tlf./Fax: (+34) 918 03 92 51 La información contenida en este mensaje y en sus archivos adjuntos es CONFIDENCIAL y se dirige exclusivamente a sus destinatarios. Queda expresamente prohibida la utilización de la misma por cualquier persona distinta de los destinatarios de esta comunicación. Si usted ha recibido este mensaje por error le rogamos que lo comunique inmediatamente a WIRELESS MUNDI y lo borre al igual que todos sus documentos adjuntos. El correo electrónico no puede asegurar la confidencialidad ni la integridad de sus mensajes por lo que WIRELESS MUNDI no se hace responsable de tales errores u omisiones. --0-- All information in this message and its attachments is confidential and may be legally privileged. Only intended recipients are authorized
Re: [asterisk-users] zaptel huge irq problem
Hi, Francois: François Delawarde wrote: Hello, I had noticed strange crackling sound on my phone calls going through my zaptel device (TDM400P), so i decided to check on possible timer issue, and found lots of issues on forums concerning the sensibility of zaptel with IRQs, and tried about everything: moving PCI slots, noapic and acpi=off boot options, play with different kernel options: iosched/preemption/timer/..., play with BIOS PCI options, change priorities, PCI latencies, IRQ balance, smp_afinity, but impossible to come up with anything correcting that problem. What kind of motherboard do you have? - The card DOES NOT seem to share interrupts (checked also with lspci): # cat /proc/interrupts CPU0 CPU1 1: 1626 0Phys-irq i8042 6: 3 0Phys-irq floppy 8: 0 0Phys-irq rtc 9: 0 0Phys-irq acpi 14: 63 0Phys-irq ide0 16: 1 0Phys-irq libata, eth3 17:6762583 0Phys-irq libata 18: 13789 0Phys-irq libata 19: 33459690 0Phys-irq eth1 20: 19864325 0Phys-irq sky2, eth0 21: 269250881 0Phys-irq wctdm 256: 77735119 0 Dynamic-irq timer0 257:3986325 0 Dynamic-irq resched0 258: 37 0 Dynamic-irq callfunc0 259: 04652748 Dynamic-irq resched1 260: 0139 Dynamic-irq callfunc1 261: 0 28924306 Dynamic-irq timer1 262: 1021 0 Dynamic-irq xenbus I've never seen cat /proc/interrupts output that looks like that... waaaitaminute... are you running this in a virtual machine? Or on a machine running virtual machines? -Stephen- ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
François Delawarde wrote: Thanks Michael, I've already been through all that unfortunately, and I have a SATA drive, so no UDMA mode 2 as far as I know. I'm currently trying everything again anyway, but i doubt it will work if nothing worked the first time. Anyone would know of issues with XEN or SMP (or both) kernel? Do dual core AMD64 processors have issues? Uh, yeah... Xen has many, many problems with interrupt handling and is utterly unsuitable for running anything that depends on hardware peripherals. I speak from very painful experience. There is no way, under any circumstance, that I would try to run Asterisk with interface cards in a Xen environment. It's too bad you wasted so much time trying to fix it -- it's never going to work. Try ripping Xen out and doing it directly on the physical server. I think you'll find your problems will go away. -Stephen- ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
Re: [asterisk-users] zaptel huge irq problem
Stephen Bosch wrote: # cat /proc/interrupts CPU0 CPU1 1: 1626 0Phys-irq i8042 6: 3 0Phys-irq floppy 8: 0 0Phys-irq rtc 9: 0 0Phys-irq acpi 14: 63 0Phys-irq ide0 16: 1 0Phys-irq libata, eth3 17:6762583 0Phys-irq libata 18: 13789 0Phys-irq libata 19: 33459690 0Phys-irq eth1 20: 19864325 0Phys-irq sky2, eth0 21: 269250881 0Phys-irq wctdm 256: 77735119 0 Dynamic-irq timer0 257:3986325 0 Dynamic-irq resched0 258: 37 0 Dynamic-irq callfunc0 259: 04652748 Dynamic-irq resched1 260: 0139 Dynamic-irq callfunc1 261: 0 28924306 Dynamic-irq timer1 262: 1021 0 Dynamic-irq xenbus I've never seen cat /proc/interrupts output that looks like that... waaaitaminute... are you running this in a virtual machine? Or on a machine running virtual machines? It looks like a XEN machine. Well spotted, Stephen. /Per Jessen, Zürich -- ENIDAN Technologies GmbH - managed email security. Starting at SFr1/month/user - http://www.spamchek.ch/ ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users
RE: [asterisk-users] zaptel huge irq problem
Try switching to a Sangoma card. You wont have anymore IRQ issues once you abandon Digium hardware. -- Salvatore Giudice [EMAIL PROTECTED] VoIP Security Training, LLC http://VoIPSecurityTraining.com 848 N. Rainbow Blvd. #1676 Las Vegas, NV 89107 Phone: (617) 959-7625 Fax: (214) 279-2906 From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of François Delawarde Sent: Monday, May 14, 2007 12:43 PM To: Asterisk Users Mailing List - Non-Commercial Discussion Subject: Re: [asterisk-users] zaptel huge irq problem Thanks Michael, I've already been through all that unfortunately, and I have a SATA drive, so no UDMA mode 2 as far as I know. I'm currently trying everything again anyway, but i doubt it will work if nothing worked the first time. Anyone would know of issues with XEN or SMP (or both) kernel? Do dual core AMD64 processors have issues? François. Michael L. Young wrote: François, I too had a similar problem and found the information on this page helpful: http://www.voip-info.org/wiki/view/Asterisk+PCI+bus+Troubleshooting What ended up working for me was changing the UDMA to mode 2 for the hard drive. Once I did that, this card has worked perfectly for me. Michael L. Young -Original Message- From: [EMAIL PROTECTED] [mailto:asterisk-users- [EMAIL PROTECTED] On Behalf Of François Delawarde Sent: Monday, May 14, 2007 10:24 AM To: asterisk-users@lists.digium.com Subject: [asterisk-users] zaptel huge irq problem Hello, I had noticed strange crackling sound on my phone calls going through my zaptel device (TDM400P), so i decided to check on possible timer issue, and found lots of issues on forums concerning the sensibility of zaptel with IRQs, and tried about everything: moving PCI slots, noapic and acpi=off boot options, play with different kernel options: iosched/preemption/timer/..., play with BIOS PCI options, change priorities, PCI latencies, IRQ balance, smp_afinity, but impossible to come up with anything correcting that problem. Any idea about this? Is it possible to force the timer to ztdummy (RTC timer) when you have a zap card plugged in? It's the only thing i could try to make it work. Thanks, François. Just in case: - Linux 2.6.18 with debian patches and xen enabled, asterisk running on dom0. - Here is my zttest results under a bit of load: # ./zttest Opened pseudo zap interface, measuring accuracy... 99.609375% 99.609375% 99.218750% 99.316406% 99.804688% 99.414062% 99.121094% 99.511719% 99.121094% 99.316406% 99.707031% 99.707031% 98.730469% 99.414062% 99.902344% 99.218750% 100.00% 99.414062% 98.828125% 99.218750% 99.316406% 98.449707% 100.00% - The card DOES NOT seem to share interrupts (checked also with lspci): # cat /proc/interrupts CPU0 CPU1 1: 1626 0Phys-irq i8042 6: 3 0Phys-irq floppy 8: 0 0Phys-irq rtc 9: 0 0Phys-irq acpi 14: 63 0Phys-irq ide0 16: 1 0Phys-irq libata, eth3 17:6762583 0Phys-irq libata 18: 13789 0Phys-irq libata 19: 33459690 0Phys-irq eth1 20: 19864325 0Phys-irq sky2, eth0 21: 269250881 0Phys-irq wctdm 256: 77735119 0 Dynamic-irq timer0 257:3986325 0 Dynamic-irq resched0 258: 37 0 Dynamic-irq callfunc0 259: 04652748 Dynamic-irq resched1 260: 0139 Dynamic-irq callfunc1 261: 0 28924306 Dynamic-irq timer1 262: 1021 0 Dynamic-irq xenbus 263: 0 0 Dynamic-irq console NMI: 0 0 LOC: 0 0 ERR: 0 MIS: 0 ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users -- _ François Delawarde Ingeniero de red Tel: 918.03.92.51 E-mail: [EMAIL PROTECTED] _ WIRELESS MUNDI http://www.wirelessmundi.com/ C/Isaac Newton, 1 - Oficina 26 · Parque Tecnológico de Madrid 28760 TRES CANTOS (Madrid) Tlf./Fax: (+34) 918 03 92 51 _ La información contenida en este mensaje y en sus archivos adjuntos es CONFIDENCIAL y se dirige exclusivamente a sus destinatarios. Queda expresamente prohibida la utilización de la misma por cualquier persona distinta de los destinatarios de esta
Re: [asterisk-users] zaptel huge irq problem
You didn't even read the thread before replying. And for what it is worth, we at Digium are very anxious to solve any sort of IRQ problems that you (or others) might have. Matthew Fredrickson On May 14, 2007, at 1:43 PM, Salvatore Giudice wrote: Try switching to a Sangoma card. You won’t have anymore IRQ issues once you abandon Digium hardware. -- Salvatore Giudice [EMAIL PROTECTED] VoIP Security Training, LLC http://VoIPSecurityTraining.com 848 N. Rainbow Blvd. #1676 Las Vegas, NV 89107 Phone: (617) 959-7625 Fax: (214) 279-2906 From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of François Delawarde Sent: Monday, May 14, 2007 12:43 PM To: Asterisk Users Mailing List - Non-Commercial Discussion Subject: Re: [asterisk-users] zaptel huge irq problem Thanks Michael, I've already been through all that unfortunately, and I have a SATA drive, so no UDMA mode 2 as far as I know. I'm currently trying everything again anyway, but i doubt it will work if nothing worked the first time. Anyone would know of issues with XEN or SMP (or both) kernel? Do dual core AMD64 processors have issues? François. Michael L. Young wrote: François, I too had a similar problem and found the information on this page helpful: http://www.voip-info.org/wiki/view/Asterisk+PCI+bus+Troubleshooting What ended up working for me was changing the UDMA to mode 2 for the hard drive. Once I did that, this card has worked perfectly for me. Michael L. Young -Original Message- From: [EMAIL PROTECTED] [mailto:asterisk-users- [EMAIL PROTECTED] On Behalf Of François Delawarde Sent: Monday, May 14, 2007 10:24 AM To: asterisk-users@lists.digium.com Subject: [asterisk-users] zaptel huge irq problem Hello, I had noticed strange crackling sound on my phone calls going through my zaptel device (TDM400P), so i decided to check on possible timer issue, and found lots of issues on forums concerning the sensibility of zaptel with IRQs, and tried about everything: moving PCI slots, noapic and acpi=off boot options, play with different kernel options: iosched/preemption/timer/..., play with BIOS PCI options, change priorities, PCI latencies, IRQ balance, smp_afinity, but impossible to come up with anything correcting that problem. Any idea about this? Is it possible to force the timer to ztdummy (RTC timer) when you have a zap card plugged in? It's the only thing i could try to make it work. Thanks, François. Just in case: - Linux 2.6.18 with debian patches and xen enabled, asterisk running on dom0. - Here is my zttest results under a bit of load: # ./zttest Opened pseudo zap interface, measuring accuracy... 99.609375% 99.609375% 99.218750% 99.316406% 99.804688% 99.414062% 99.121094% 99.511719% 99.121094% 99.316406% 99.707031% 99.707031% 98.730469% 99.414062% 99.902344% 99.218750% 100.00% 99.414062% 98.828125% 99.218750% 99.316406% 98.449707% 100.00% - The card DOES NOT seem to share interrupts (checked also with lspci): # cat /proc/interrupts CPU0 CPU1 1: 1626 0 Phys-irq i8042 6: 3 0 Phys-irq floppy 8: 0 0 Phys-irq rtc 9: 0 0 Phys-irq acpi 14: 63 0 Phys-irq ide0 16: 1 0 Phys-irq libata, eth3 17: 6762583 0 Phys-irq libata 18: 13789 0 Phys-irq libata 19: 33459690 0 Phys-irq eth1 20: 19864325 0 Phys-irq sky2, eth0 21: 269250881 0 Phys-irq wctdm 256: 77735119 0 Dynamic-irq timer0 257: 3986325 0 Dynamic-irq resched0 258: 37 0 Dynamic-irq callfunc0 259: 0 4652748 Dynamic-irq resched1 260: 0 139 Dynamic-irq callfunc1 261: 0 28924306 Dynamic-irq timer1 262: 1021 0 Dynamic-irq xenbus 263: 0 0 Dynamic-irq console NMI: 0 0 LOC: 0 0 ERR: 0 MIS: 0 ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users ___ --Bandwidth and Colocation provided by Easynews.com -- asterisk-users mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users -- _ François Delawarde Ingeniero de red Tel: 918.03.92.51 E-mail: [EMAIL PROTECTED] _ WIRELESS MUNDI http://www.wirelessmundi.com/ C/Isaac Newton, 1 - Oficina 26 · Parque Tecnológico de Madrid 28760 TRES CANTOS (Madrid) Tlf./Fax: (+34) 918 03 92 51 La