Hello, sry for HTML Mail, it is auto configuration of Outlook.
We have testet adding device_model_version="qemu-xen" to the config and rebooting the VMs. With this configuration, no change at memory usage. Mit freundlichen Grüßen Michael Schinzel - Geschäftsführer - IP-Projects GmbH & Co. KG Am Vogelherd 14 D - 97295 Waldbrunn Telefon: 09306 - 76499-0 FAX: 09306 - 76499-15 E-Mail: i...@ip-projects.de Geschäftsführer: Michael Schinzel Registergericht Würzburg: HRA 6798 Komplementär: IP-Projects Verwaltungs GmbH -----Ursprüngliche Nachricht----- Von: Paul Durrant [mailto:paul.durr...@citrix.com] Gesendet: Donnerstag, 31. August 2017 10:44 An: Michael Schinzel <schin...@ip-projects.de>; xen-devel@lists.xen.org Cc: Thomas Toka <t...@ip-projects.de> Betreff: RE: Memory Issue HVM guest after Upgrade from 4.4 to 4.8 De-htmling... My response indented... From: Xen-devel [mailto:xen-devel-boun...@lists.xen.org] On Behalf Of Michael Schinzel Sent: 31 August 2017 08:43 To: xen-devel@lists.xen.org Cc: Thomas Toka <t...@ip-projects.de> Subject: [Xen-devel] Memory Issue HVM guest after Upgrade from 4.4 to 4.8 Hello, cause of the not longer support of xen 4.4 hypervisor, we actually upgrade all of our xen hosts from 4.4 - debian 8 to 4.8 - debian 9. Till this update, we run a host for example with 16 GB memory for dom0 with about 93 VMs. Till the upgrade, all memory usage was fine. The Host use about 1.4 - 3 GB memory of the allocated 16 GB. At each host, we mix HVM and Para VMs. After the Upgrade, the HVM VMs constantly use more and more memory. About 100 MB more each 2-3 Minutes until the Host swaps. The Problem is only with HVM VMs, Para is all fine. top - 09:40:43 up 1 day, 2:28, 1 user, load average: 0,94, 1,03, 1,27 Tasks: 1313 total, 5 running, 1308 sleeping, 0 stopped, 0 zombie %Cpu(s): 0,2 us, 0,8 sy, 0,0 ni, 98,3 id, 0,0 wa, 0,0 hi, 0,3 si, 0,5 st KiB Mem : 30315388 total, 10791316 free, 18483368 used, 1040704 buff/cache KiB Swap: 15998972 total, 15919092 free, 79880 used. 11492116 avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 17962 root 20 0 1202580 680040 17356 R 5,2 2,2 1:20.06 qemu-system-i38 21679 root 20 0 1454316 922580 17760 R 5,2 3,0 1:26.99 qemu-system-i38 27772 root 20 0 1635108 1,082g 17524 S 5,2 3,7 1:27.50 qemu-system-i38 29731 root 20 0 1374844 896052 17016 S 5,2 3,0 1:17.82 qemu-system-i38 14209 root 20 0 1130476 597120 17724 S 4,9 2,0 1:17.24 qemu-system-i38 19846 root 20 0 1417076 921928 16952 S 4,6 3,0 1:31.96 qemu-system-i38 4830 root 20 0 2092624 1,496g 17640 S 3,9 5,2 1:53.88 qemu-system-i38 18897 root 20 0 2013120 1,353g 17932 S 3,9 4,7 1:30.63 qemu-system-i38 7832 root 20 0 46296 5160 3176 R 2,3 0,0 0:00.38 top 31832 root 20 0 1373044 835140 17688 S 2,3 2,8 0:48.82 qemu-system-i38 28744 root 20 0 1053868 530680 17944 S 2,0 1,8 0:34.11 qemu-system-i38 13307 root 20 0 913684 424984 17168 S 1,6 1,4 0:27.57 qemu-system-i38 15248 root 20 0 1411316 887892 17608 S 1,6 2,9 0:43.97 qemu-system-i38 16135 root 20 0 1204240 640644 17776 S 1,6 2,1 0:37.28 qemu-system-i38 20763 root 20 0 1036288 513848 17484 S 1,6 1,7 0:35.76 qemu-system-i38 22663 root 20 0 851712 301236 17588 S 1,6 1,0 0:28.24 qemu-system-i38 24849 root 20 0 1164908 644736 17824 S 1,6 2,1 0:39.93 qemu-system-i38 25871 root 20 0 1113684 571548 17616 S 1,6 1,9 0:37.14 qemu-system-i38 26840 root 20 0 1045604 515216 17888 S 1,6 1,7 0:35.42 qemu-system-i38 30693 root 20 0 2329944 1,734g 17644 S 1,6 6,0 1:32.11 qemu-system-i38 23743 root 20 0 1470544 929708 17500 S 1,3 3,1 0:47.23 qemu-system-i38 The config file of one HVM guest: #kernel = "hvmloader" builder='hvm' memory = 512 maxmem = 512 shadow_memory = 8 name = "vmanager1157" vif = [ 'vifname=vmanager1157, rate=100Mb/s, bridge=xenbr0.240, mac=xxx, ip=xxx 2001:1608:10:3:0:0:c:1' ] vif_other_config = [ 'xxx, 'tbf', 'rate=100Mb/s', 'bps_read=150Mb/s', 'bps_write=150Mb/s', 'iops_read=150000IOPS', 'iops_write=150000IOPS' ] disk = [ 'phy:/dev/vm/vmanager1157-root,xvda,w', 'file:/root/vmanager/iso/CentOS-7.0-1406-x86_64-NetInstall.iso,xvdc:cdrom,r' ] boot="cd" vcpus = 1 sdl=0 vnc=1 vnclisten="0.0.0.0" vncdisplay=69 vncpasswd='3s2Xwv65' vncunused=0 stdvga=0 serial='pty' usbdevice='tablet' on_poweroff = 'destroy' on_reboot = 'restart' on_crash = 'destroy' Is this normal with the new Xen Hypervisor? Actually we use Kernel Version 4.12.10. So the newest Kernel with Xen Support. > The default choice for QEMU changed between 4.4 and 4.8. In 4.4 is was trad > and in 4.8 it is upstream. If you force use of trad in your config, do you > still see the apparent leak? > > Paul Yours sincerely Michael Schinzel _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org https://lists.xen.org/xen-devel
