hugepage_reset: Test compatible with different NUMA topologies #4237
Conversation
|
Please @JinLiul could you test this PR whenever you have a 8 NUMA nodes system again? Thanks! |
|
Hi @mcasquer, tested with 8 NUMA nodes system. |
mcasquer
force-pushed
the
3254_hp_reset_setup
branch
from
5e1c2c5 to
b1013a9
Compare
mcasquer
changed the title
|
Tests results on a 8 NUMA nodes host (with test loop a bit tuned 😁) |
|
Please @JinLiul could you test again this PR? Thanks ! |
|
Also passed in 2 NUMA nodes host |
|
huge_page_1G test loop passed |
mcasquer
force-pushed
the
3254_hp_reset_setup
branch
from
b1013a9 to
d77d567
Compare
| @@ -107,9 +125,9 @@ def heavyload_install(): | |||
| "No node on your host has sufficient free memory for " "this test." | |||
| ) | |||
| hp_config = test_setup.HugePageConfig(params) | |||
| if params.get("on_numa_node"): | |||
| allocate_largepages_per_node() | |||
There was a problem hiding this comment.
@mcasquer , code LGTM, I just want to confirm with you that if the node mem is not enough, still setup or better to raise error or skip the test.
There was a problem hiding this comment.
Mmmm @PaulYuuu good point, I think that situation should be handled, perhaps with a try block, I'll send an update of this
There was a problem hiding this comment.
@PaulYuuu added a try block that will cancel the case ig there's no enough memory, faked example:
(2/2) Host_RHEL.m9.u6.ovmf.qcow2.virtio_scsi.up.virtio_net.Guest.RHEL.9.6.0.x86_64.io-github-autotest-qemu.hugepage_reset.on_numa_node.q35: STARTED
(2/2) Host_RHEL.m9.u6.ovmf.qcow2.virtio_scsi.up.virtio_net.Guest.RHEL.9.6.0.x86_64.io-github-autotest-qemu.hugepage_reset.on_numa_node.q35: CANCEL: 18 (expecting 400) hugepages is set on the node 0, please check if the node has enough memory (12.51 s)
As the test will set 8 hugepages, this works fine for systems with 2 NUMA nodes, having e.g. 8 nodes is going to lead the on_numa_node variant to fail since the binded node doesn't have enough hugepages. As the cfg already suggests to allocate 1G hugepages on boot time, let's make user decision how many hugepages allocate, adding an informative comment in the cfg as well. Finally, if system hugepage_size is 1GB, allocates at runtime enough hugepages in all valid nodes. Signed-off-by: mcasquer <[email protected]>
mcasquer
force-pushed
the
3254_hp_reset_setup
branch
from
d77d567 to
3b3e2de
Compare
hugepage_reset: Test compatible with different NUMA topologies
As the test will set 8 hugepages, this works fine
for systems with 2 NUMA nodes, having e.g. 8 nodes
is going to lead the on_numa_node variant to fail
since the binded node doesn't have enough hugepages.
As the cfg already suggests to allocate 1G hugepages
on boot time, let's make user decision how many hugepages
allocate, adding an informative comment in the cfg as well.
Finally, if system hugepage_size is 1GB, allocates at
runtime enough hugepages in all valid nodes.
Signed-off-by: mcasquer [email protected]
ID: 3254