From: "Massimo S." Received: from [192.168.100.201] (HELO mail.2rosenthals.com) by 2rosenthals.com (CommuniGate Pro SMTP 5.4.10) with ESMTPS id 310836 for virtualized_ecs_users@2rosenthals.com; Tue, 28 Jan 2020 15:44:42 -0500 Received: from [192.168.200.201] (port=48196 helo=mail2.2rosenthals.com) by mail.2rosenthals.com with esmtp (Exim 4.82_1-5b7a7c0-XX) (envelope-from ) id 1iwXjd-0008NA-0v for virtualized_ecs_users@2rosenthals.com; Tue, 28 Jan 2020 15:45:21 -0500 Received: from mail2.quasarbbs.net ([80.86.52.115]:10049) by mail2.2rosenthals.com with esmtp (Exim 4.82_1-5b7a7c0-XX) (envelope-from ) id 1iwXej-0002Rh-3C for virtualized_ecs_users@2rosenthals.com; Tue, 28 Jan 2020 15:40:18 -0500 Received: from [192.168.10.199] (dtp [192.168.10.199]) by srv2 (Weasel v2.71) for ; 28 Jan 2020 21:40:14 X-CTCH-RefID: str=0001.0A02020A.5E309C32.0049,ss=1,re=0.000,recu=0.000,reip=0.000,cl=1,cld=1,fgs=0 Reply-To: ml@ecomstation.it Subject: Re: [Virtualized eCS] Virtual box 6.0.14 r 133895, acpi shutdown with AOS 5.0.3 and virtual cpu - 2 cpu success - OK!!! :) To: Virtualized eCS Users Mailing List References: Organization: eComStation dot it Message-ID: <48237de9-71c6-fdd3-8a9b-4fd586e60641@ecomstation.it> Date: Tue, 28 Jan 2020 21:40:12 +0100 User-Agent: Mozilla/5.0 (OS/2; U; Warp 4.5; it-IT; rv:1.7.13) Gecko/20060424 Thunderbird/1.0.8 Mnenhy/0.7.4.0 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=utf-8; format=flowed Content-Language: it-IT Content-Transfer-Encoding: 7bit Il 28/01/2020 20:06, Doug Bissett ha scritto: > On 2020-01-28, at 10:01:57, Massimo S. wrote: >> >> >> >> Il 28/01/2020 04:56, Doug Bissett ha scritto: > > ...snip... > >>> I don't know anything about any of them. I use VMs to be able to > run >>> various versions of software, so I can try things, when something >>> isn't working. At the moment, I don't have a good setup to try >>> running a FTP server, so I can't try that without doing another >>> install. >> >> it's a different kind of use, you run guests for lab purposes, >> for devs use, or test use > > True... > >> i must run VMs for server use (web, email, ftpd, webmail etc..) >> so they must run 24 hours per day >> and they have to be: >> >> 1) stable and reliable >> 2) fast in performances more or less like on the bare metal >> 3) they must bkup theirself without having to be powered off >> >> point 1 and 3 we are OK now with Vbox >> the bad point is the number 2 and only for LAN speed > > It would be interesting to find the cause. It should be just as fast as > bare metal, but there are so many factors involved, that it could be > difficult to find out why it is slower. > >>>> >>>>>> if i find a free sw for live bkups of VMs under ESX >>>>>> i will go with ESX has hypervisor >>>>>> >>>>>> massimo >>>> >From the little that I know about ESX, it does not virtualize a >>>>> machine. It shares a real machine with more than one > operating >>>>> system. If it works, use it. >>>> >>>> i still have to find how to schedule bkups >>>> how to add notifications >>>> >>>> i've found "GhettoVCB"* that's free runs under ESX free >>> hypervisor >>>> and do live bkups (i mean bkups without turning off the VMs) >>>> but i still have to find a way to schedule it >>>> e.g. one time per night... >>> I never bothered under windows, although the Task Scheduler >>> should run a program, at a scheduled time. Most operating > systems >>> have some sort of Cron program. >> >> you are a bit confused >> Vmware Vsphere ESX Hypervisor (free, but also the versions not > free) >> has its own OS that runs on the bare metal and it's not Windows >> >> ( while MS Hyper-V runs on Windows core, a free version of >> Windows server that comes without the GUI, but has a free >> and very powerful offline web gui, it's name is "WAC" >> https://docs.microsoft.com/it-it/windows-server/manage/windows- > admin-cen >> ter/overview ) >> >> maybe noone knows here, but there is a Windows server free >> it's name is Windows server core, it has no windows gui >> but you can admin it with batchs, powershell and WAC > > Interesting. Windows, that isn't windows. I can't think of any > possible use for something like that, but it is good to know about > these things. > >> >>>> and it uses NFS as protocol that i don't like so much >>>> anyway i also have a Qnap nap that use NFS >>>> so i could use it as storage for bkups... >>>> >>>> * https://communities.vmware.com/docs/DOC-8760 >>> I don't use windows enough to worry about nightly backups. >> again VMware vsphere esx hypervisor don't runs on Windows >> it's a CD that you install on bare metal and you have >> a web gui to administer it >> you can also activate an ssh with a shell that's similar to *nix >> >>> I use >>> RSync under OS/2 (using the DragText scheduler). There is >>> DeltaCopy, that runs RSync under windows. I don't know if it > would >>> do what you want. >> >> i use rsync under eCS to bkup application servers like mail, web > server >> etc.. >> >> all theese kind of application servers bkups will stay also on my > new >> Virtualized architecture, but i also want the full bkup of running > VMs >> i don't want to have to reinstall eCS or AOS if a server die > > It seems that multiple backups, using different methods, on different > types of backup hardware, is a good idea, even when a RAID is > used. It was the RAID unit, that took out everything, when Arca > Noae lost their servers. > >> i want to simply import the VMs bkups on the server 2 and play > the VMs > > I know that VBox will Export a complete VM to a file, and it can > then Import that file to recreate the VM. I have done that a few > times, and it seems to work. I don't know if it will Export a running > VM (probably not). > >>>>> I still think you have a bad setup in VBox, but I don't have the >>> time >>>>> to try to figure it out. >>>> the CPU is a xeon 8cores but clocked at 2Ghz, i guess that > Vbox >>> is very >>>> "clock/frequency" oriented, the much higher is the clock, the > faster >>> are >>>> performances >>> There are a lot of pieces that fit together, to improve > performance. >>> A fast host definitely helps, but that isn't the only thing. >> >> yes, but compeared with MS Hyper-V or Vmware Vsphere, Vbox >> it seems very limited to me... > > It is limited, but it should work near full speed, for what you are > doing. Something is wrong. > > I have an old P4 processor, with 1 GB of memory, that I use for > much the same things that you mention. Network speed is not an > issue, although there are a few things that I haven't figured out. For > instance, up until about a year ago, windows, using a SAMBA > connection to copy a large file to the server, claimed 20 MBs (on a > Gbit network), and OS/2 (the P4) claimed 16 Mbs. There was a > windows update, and windows started claiming 16 Gbs, which > agreed with OS/2. That seems to be as fast as that connection will > go. However, If I use another windows machine, to connect to the > P4, at the same time, they both go at about 15 Mbs (the P4 then > shows 30 Mbs). If I add another windows machine, they all run at > about 13 Mbs, and the P4 shows about 39 Mbs. My conclusion is, > that windows is governing the speed, somehow, because OS/2 can > certainly go much faster than any individual connection from > windows. I do see similar numbers, when I use OS/2 instead of > windows, so there is something else involved. > >> >>>> this is my tought at the moment >>>> as far from my tests, MS free Hyper-V is way the faster >>> hypervisor >>>> around >>>> but unfortunately eCS/AOS don't run under it :( >>>> this one does everything, live bkups, live replica and >>>> comes with a number of interesting features >>> AFAIK, Hyper-V only works with windows guests. >> >> absolutely not, this is a fake > > As I mentioned, I don't know much about Hyper-V. > >> Hyper-V is tested, used and certified to run also >> a number of linux distros (redhat, ubuntu, suse etc...) >> >> old versions of HV also ran eCS2.2b, but after some >> update (i guess version 8 of VMs it's no more compatible) > > It would not surprise me, at all, to find out that it was programmed to > recognize OS/2, and refuse to run it, just because it is OS/2. > >>>>> there was a post, some time ago, in this mail >>>>> list, about how to make a modern VBox setup for OS/2 >>> (probably >>>>> eCS at the time) >>>> >>>> i'm subscribed since 3th jan 2020 :( >>>> >>>> >>>> massimo >>> Web archives are publicly available at: > http://lists.2rosenthals.com, >>> but it seems that a lot of content went missing, when the Arca > Noae >>> servers crashed. I keep them for one year, then delete them. > What >>> you need would be somewhere in the middle of the missing stuff. >>> I guess I need to do the setup thing all over again, when I find the >>> time. >> >> anyway let's try to understand if there is something no good >> let's take VM1: >> >> general: >> type: IBM OS/2 >> version: eComStation > > That is okay for eCS (based on MCP2). ArcaOS is closer to Warp > 4.5 (based on warp server). I don't think there is much difference, > but it could make wrong assumptions. > >> advanced: shared clipboard: shared > > Try these disabled, to see if anything changes. I have heard > comments about problems with that. I never enable them. > >> system: >> memory: 4096 MB > > Probably way too much (that will also slow down the host). In fact, > OS/2 can only use somewhere near 3500 MB, depending on the > exact machine configuration. I would suggest using 2000 MB (not > even 2 GB). I find, with my P4 server, which has 1 GB of memory, > that it NEVER uses that much memory. In fact I have never seen it > use more than about 600 MB. > >> boot order: CD and HDD >> chipset: PIIX3 >> point dev: PS/2 mouse >> extended functions: I/O APIC on >> processors: 2 > > I am still surprised that this doesn't cause problems. > >> execution cap: 100% >> estended features: PAE/NX: on turned OFF > AH!, Now this might cause problems. The only thing, in OS/2, that > can use PAE, is the QSINIT RAMDISK (the ArcaOS loader, is > QSINIT, and the ArcaOS RAMDISK is from QSINIT). Some time > ago, the guy who wrote that software has said that there are > problems with PAE/NX in VBox. That may have changed, but I > doubt it. Try turning it off. > >> Vt-x/AMD-V nidificate: off > > This setting will be disabled, when you turn off PAE/NX. > >> acceleration: >> default set to legacy >> VT-x/AMD-V: on > > I don't see this in the VBox 6.1.2 setup, any more. > Paravirtualization Interface should be set to Legacy, which would > seem to be the same thing. Of course, VT-x/AMD-V must be > enabled in the BIOS too. > >> paging nidificate: off set to on > Nested paging can (should) be on. This means that the host file > system can cache the file that the guest uses as a hard disk. It may > not make a lot of difference, if the host disk is a SSD. yes, SSD and the OS is on a different hard-disk (2 raid1 hdds) while the 2 VMs are on a SSD (kingston ssdnow 300v) the production machine will run on samsung evo pro even better SSDs >> screen: >> all default values > > Default should be okay, most of the time. I usually change Video > Memory to 15 or 24 MB, depending on how big the host screen is. It > usually insists on more, if it needs more memory to run the guest in > full screen. Apparently, it doesn't actually allocate the memory, > unless it is needed. > >> storage: IDE/PIIX4 > > Hmmm. This should be AHCI (but be sure the OS2AHCI driver is > configured in CONFIG.SYS, before changing it). CONFIG.SYS > should contain: > BASEDEV=OS2AHCI.ADD /n /f > Put it before the DANI driver line, if you leave both in CONFIG.SYS > (recommended, when changing it). as wrote before if i switch to OS2ACHI + sata i see a tremendous drop in perfomances on the files systems inside the VMs >> use host cache I/O: yes >> type of storage: vmdk file > > I use VDI, but VMDK is okay. vmdk is the "raw" format must be use in the case you need (like me) live bkups if you use Windows server the service VSS and vmdk as vm format permit to bkup VMs live when they are running >> ide primary master > > Change that to SATA, so it can work as AHCI. > >> ssd: yes > > You might try turning this off. I doubt if it will help, but it is > something to try. > >> audio: disabled > > If you do use audio, select the Sound Blaster 16 PNP. > >> lan: >> >> my architecture need 2 NICs per VM >> >> nic 1 >> connection: NIC with bridge >> name: Intel I350 Gigabit Network COnnection >> advanced: >> type: INtel PRO/1000 MT server (82545EM) >> mode: permit all >> mac .... >> cable: connected >> >> nic 2 >> connection: NIC with bridge >> name: Intel I350 Gigabit Network COnnection 2 >> advanced: >> type: INtel PRO/1000 MT server (82545EM) >> mode: permit all > > I don't know what mode really does. Promiscuous mode is used to > allow a NIC to be used by something like VBox, but you can't run a > VBox inside of a VBox. I would try turning that off. > > I don't know, exactly, how you connect the networks, but OS/2 > cannot connect to two networks that are in the same IP address > range. they have different IP classes ...Doug, thanks a lot!! here we go, now i see via ftp a constat datarate of about 20MegaBytes/sec this not wonderful, but *it's enough for my needs* ok, WS2019+Virtual Box has won, vmware esx free has lost (too much complicated too manage, a number of issues in the web interface too, difficult to schedule bkup tasks) ;-) massimo > >> mac ....... >> cable: connected >> >> serial ports: all disabled >> usb: all disabled >> >> that's all >> >> the VMs are stored on a single Kingston SSD >> (in the future will run on a raid1-mirroring with Samsung evo SSD) >> >> if i increase the core number of the guest VM, the lan speed (ftp >> datarates improves) > > I think your biggest slow down, is using IDE rather than > SATA/AHCI. Promiscuous mode, on the NICs, may affect network > speed, and would not be used by anything in OS/2. I do know that > having it on will force the NIC to handle every network packet that > goes by, whether it is for that machine, or not (meaning the VM). > When it is off, it will just ignore any packet that goes by, unless it is > for that machine. The host needs to have promiscuous mode on, so > that VBox can use it. > > As I said, it all works together, and any one of those things could > cause a performance problem. >