Re: [libvirt] [PATCH 00/11] Allow modification of IOThread polling values (redux)
On 10/15/2018 04:28 PM, John Ferlan wrote: > > ping? > > Any takers or thoughts? No review, but I think it makes a lot of sense to expose these tuneables. > > Tks, > > John > > > On 10/7/18 9:00 AM, John Ferlan wrote: >> This series attempts to resurrect the concept of being able to modify >> the IOThread polling parameters; however, in a slightly different >> manner than the previous series posted by posted by Pavel Hrdina >> : >> >> https://www.redhat.com/archives/libvir-list/2017-February/msg01047.html >> >> The work is prompted by continued pleas found in the bz: >> >> https://bugzilla.redhat.com/show_bug.cgi?id=1545732 >> >> to provide some way to modify the paremters without needing to supply >> QEMU command line pass through values. >> >> It's accepted that the values being changed are fairly or extremely >> low level type knobs; however, it's also shown that by being able to >> turn the knob it is possible for certain, specific appliances to be >> able to gain a performance benefit for the thread at the expense of >> other competing threads. >> >> Unlike the previous series, this series does not attempt to save the >> polling values in the guest XML. Rather, it only modifies the live >> guest's IOThread with new values. It also doesn't provide the polling >> values in a virsh iothread* command, rather it uses the domstats >> in order to fetch and display the values. The theory being this >> leaves the onus on the higher level appliance/application to provide >> the "proper guidance" and "concerns" related to changing the values >> to the consumer. Not saving the values means whatever values that >> are chosen do not "live" in perpetuity. Once the guest is shut down >> or the IOThread removed from guest, the hypervisor default values >> take over again. Perhaps not a perfect situation in terms of what >> the bz requests; however, storage of default values that could >> cause performance issues is not an optimal situation. So this I >> figured is a "comprimise" of sorts. >> >> If it's still felt that no we don't want to do this, then fine, >> but please in doing so own the bz, state your case, and close it. >> I'm 50/50 on it, but figured at least I'd present this option and >> see what the concensus was. >> >> John Ferlan (11): >> qemu: Check for and return IOThread polling values if available >> qemu: Split qemuDomainGetIOThreadsLive >> qemu: Implement the ability to return IOThread stats >> virsh: Add ability to display IOThread stats >> lib: Introduce virDomainSetIOThreadParams >> qemu: Add monitor functions to set IOThread params >> qemu: Alter qemuDomainChgIOThread to take enum instead of bool >> qemu: Alter qemuDomainChgIOThread to take qemuMonitorIOThreadInfo >> qemu: Detect whether iothread polling is supported >> qemu: Introduce qemuDomainSetIOThreadParams >> tools: Add virsh iothreadset command >> >> include/libvirt/libvirt-domain.h | 45 ++ >> src/driver-hypervisor.h | 8 + >> src/libvirt-domain.c | 109 + >> src/libvirt_public.syms | 5 + >> src/qemu/qemu_capabilities.c | 2 + >> src/qemu/qemu_capabilities.h | 1 + >> src/qemu/qemu_driver.c| 393 -- >> src/qemu/qemu_monitor.c | 19 + >> src/qemu/qemu_monitor.h | 6 + >> src/qemu/qemu_monitor_json.c | 48 +++ >> src/qemu/qemu_monitor_json.h | 4 + >> src/remote/remote_driver.c| 1 + >> src/remote/remote_protocol.x | 21 +- >> src/remote_protocol-structs | 10 + >> .../caps_2.10.0.aarch64.xml | 1 + >> .../caps_2.10.0.ppc64.xml | 1 + >> .../caps_2.10.0.s390x.xml | 1 + >> .../caps_2.10.0.x86_64.xml| 1 + >> .../caps_2.11.0.s390x.xml | 1 + >> .../caps_2.11.0.x86_64.xml| 1 + >> .../caps_2.12.0.aarch64.xml | 1 + >> .../caps_2.12.0.ppc64.xml | 1 + >> .../caps_2.12.0.s390x.xml | 1 + >> .../caps_2.12.0.x86_64.xml| 1 + >> .../qemucapabilitiesdata/caps_2.9.0.ppc64.xml | 1 + >> .../qemucapabilitiesdata/caps_2.9.0.s390x.xml | 1 + >> .../caps_2.9.0.x86_64.xml | 1 + >> .../qemucapabilitiesdata/caps_3.0.0.ppc64.xml | 1 + >> .../caps_3.0.0.riscv32.xml| 1 + >> .../caps_3.0.0.riscv64.xml| 1 + >> .../qemucapabilitiesdata/caps_3.0.0.s390x.xml | 1 + >> .../caps_3.0.0.x86_64.xml | 1 + >> tools/virsh-domain-monitor.c | 7 + >> tools/virsh-domain.c | 110 + >> tools/virsh.pod | 47 ++- >> 35 files changed, 810 insertions(+
Re: [libvirt] [PATCH 00/11] Allow modification of IOThread polling values (redux)
ping? Any takers or thoughts? Tks, John On 10/7/18 9:00 AM, John Ferlan wrote: > This series attempts to resurrect the concept of being able to modify > the IOThread polling parameters; however, in a slightly different > manner than the previous series posted by posted by Pavel Hrdina > : > > https://www.redhat.com/archives/libvir-list/2017-February/msg01047.html > > The work is prompted by continued pleas found in the bz: > > https://bugzilla.redhat.com/show_bug.cgi?id=1545732 > > to provide some way to modify the paremters without needing to supply > QEMU command line pass through values. > > It's accepted that the values being changed are fairly or extremely > low level type knobs; however, it's also shown that by being able to > turn the knob it is possible for certain, specific appliances to be > able to gain a performance benefit for the thread at the expense of > other competing threads. > > Unlike the previous series, this series does not attempt to save the > polling values in the guest XML. Rather, it only modifies the live > guest's IOThread with new values. It also doesn't provide the polling > values in a virsh iothread* command, rather it uses the domstats > in order to fetch and display the values. The theory being this > leaves the onus on the higher level appliance/application to provide > the "proper guidance" and "concerns" related to changing the values > to the consumer. Not saving the values means whatever values that > are chosen do not "live" in perpetuity. Once the guest is shut down > or the IOThread removed from guest, the hypervisor default values > take over again. Perhaps not a perfect situation in terms of what > the bz requests; however, storage of default values that could > cause performance issues is not an optimal situation. So this I > figured is a "comprimise" of sorts. > > If it's still felt that no we don't want to do this, then fine, > but please in doing so own the bz, state your case, and close it. > I'm 50/50 on it, but figured at least I'd present this option and > see what the concensus was. > > John Ferlan (11): > qemu: Check for and return IOThread polling values if available > qemu: Split qemuDomainGetIOThreadsLive > qemu: Implement the ability to return IOThread stats > virsh: Add ability to display IOThread stats > lib: Introduce virDomainSetIOThreadParams > qemu: Add monitor functions to set IOThread params > qemu: Alter qemuDomainChgIOThread to take enum instead of bool > qemu: Alter qemuDomainChgIOThread to take qemuMonitorIOThreadInfo > qemu: Detect whether iothread polling is supported > qemu: Introduce qemuDomainSetIOThreadParams > tools: Add virsh iothreadset command > > include/libvirt/libvirt-domain.h | 45 ++ > src/driver-hypervisor.h | 8 + > src/libvirt-domain.c | 109 + > src/libvirt_public.syms | 5 + > src/qemu/qemu_capabilities.c | 2 + > src/qemu/qemu_capabilities.h | 1 + > src/qemu/qemu_driver.c| 393 -- > src/qemu/qemu_monitor.c | 19 + > src/qemu/qemu_monitor.h | 6 + > src/qemu/qemu_monitor_json.c | 48 +++ > src/qemu/qemu_monitor_json.h | 4 + > src/remote/remote_driver.c| 1 + > src/remote/remote_protocol.x | 21 +- > src/remote_protocol-structs | 10 + > .../caps_2.10.0.aarch64.xml | 1 + > .../caps_2.10.0.ppc64.xml | 1 + > .../caps_2.10.0.s390x.xml | 1 + > .../caps_2.10.0.x86_64.xml| 1 + > .../caps_2.11.0.s390x.xml | 1 + > .../caps_2.11.0.x86_64.xml| 1 + > .../caps_2.12.0.aarch64.xml | 1 + > .../caps_2.12.0.ppc64.xml | 1 + > .../caps_2.12.0.s390x.xml | 1 + > .../caps_2.12.0.x86_64.xml| 1 + > .../qemucapabilitiesdata/caps_2.9.0.ppc64.xml | 1 + > .../qemucapabilitiesdata/caps_2.9.0.s390x.xml | 1 + > .../caps_2.9.0.x86_64.xml | 1 + > .../qemucapabilitiesdata/caps_3.0.0.ppc64.xml | 1 + > .../caps_3.0.0.riscv32.xml| 1 + > .../caps_3.0.0.riscv64.xml| 1 + > .../qemucapabilitiesdata/caps_3.0.0.s390x.xml | 1 + > .../caps_3.0.0.x86_64.xml | 1 + > tools/virsh-domain-monitor.c | 7 + > tools/virsh-domain.c | 110 + > tools/virsh.pod | 47 ++- > 35 files changed, 810 insertions(+), 44 deletions(-) > -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
[libvirt] [PATCH 00/11] Allow modification of IOThread polling values (redux)
This series attempts to resurrect the concept of being able to modify the IOThread polling parameters; however, in a slightly different manner than the previous series posted by posted by Pavel Hrdina : https://www.redhat.com/archives/libvir-list/2017-February/msg01047.html The work is prompted by continued pleas found in the bz: https://bugzilla.redhat.com/show_bug.cgi?id=1545732 to provide some way to modify the paremters without needing to supply QEMU command line pass through values. It's accepted that the values being changed are fairly or extremely low level type knobs; however, it's also shown that by being able to turn the knob it is possible for certain, specific appliances to be able to gain a performance benefit for the thread at the expense of other competing threads. Unlike the previous series, this series does not attempt to save the polling values in the guest XML. Rather, it only modifies the live guest's IOThread with new values. It also doesn't provide the polling values in a virsh iothread* command, rather it uses the domstats in order to fetch and display the values. The theory being this leaves the onus on the higher level appliance/application to provide the "proper guidance" and "concerns" related to changing the values to the consumer. Not saving the values means whatever values that are chosen do not "live" in perpetuity. Once the guest is shut down or the IOThread removed from guest, the hypervisor default values take over again. Perhaps not a perfect situation in terms of what the bz requests; however, storage of default values that could cause performance issues is not an optimal situation. So this I figured is a "comprimise" of sorts. If it's still felt that no we don't want to do this, then fine, but please in doing so own the bz, state your case, and close it. I'm 50/50 on it, but figured at least I'd present this option and see what the concensus was. John Ferlan (11): qemu: Check for and return IOThread polling values if available qemu: Split qemuDomainGetIOThreadsLive qemu: Implement the ability to return IOThread stats virsh: Add ability to display IOThread stats lib: Introduce virDomainSetIOThreadParams qemu: Add monitor functions to set IOThread params qemu: Alter qemuDomainChgIOThread to take enum instead of bool qemu: Alter qemuDomainChgIOThread to take qemuMonitorIOThreadInfo qemu: Detect whether iothread polling is supported qemu: Introduce qemuDomainSetIOThreadParams tools: Add virsh iothreadset command include/libvirt/libvirt-domain.h | 45 ++ src/driver-hypervisor.h | 8 + src/libvirt-domain.c | 109 + src/libvirt_public.syms | 5 + src/qemu/qemu_capabilities.c | 2 + src/qemu/qemu_capabilities.h | 1 + src/qemu/qemu_driver.c| 393 -- src/qemu/qemu_monitor.c | 19 + src/qemu/qemu_monitor.h | 6 + src/qemu/qemu_monitor_json.c | 48 +++ src/qemu/qemu_monitor_json.h | 4 + src/remote/remote_driver.c| 1 + src/remote/remote_protocol.x | 21 +- src/remote_protocol-structs | 10 + .../caps_2.10.0.aarch64.xml | 1 + .../caps_2.10.0.ppc64.xml | 1 + .../caps_2.10.0.s390x.xml | 1 + .../caps_2.10.0.x86_64.xml| 1 + .../caps_2.11.0.s390x.xml | 1 + .../caps_2.11.0.x86_64.xml| 1 + .../caps_2.12.0.aarch64.xml | 1 + .../caps_2.12.0.ppc64.xml | 1 + .../caps_2.12.0.s390x.xml | 1 + .../caps_2.12.0.x86_64.xml| 1 + .../qemucapabilitiesdata/caps_2.9.0.ppc64.xml | 1 + .../qemucapabilitiesdata/caps_2.9.0.s390x.xml | 1 + .../caps_2.9.0.x86_64.xml | 1 + .../qemucapabilitiesdata/caps_3.0.0.ppc64.xml | 1 + .../caps_3.0.0.riscv32.xml| 1 + .../caps_3.0.0.riscv64.xml| 1 + .../qemucapabilitiesdata/caps_3.0.0.s390x.xml | 1 + .../caps_3.0.0.x86_64.xml | 1 + tools/virsh-domain-monitor.c | 7 + tools/virsh-domain.c | 110 + tools/virsh.pod | 47 ++- 35 files changed, 810 insertions(+), 44 deletions(-) -- 2.17.1 -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list