You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@mesos.apache.org by Meng Zhu <mz...@mesosphere.io> on 2018/02/27 00:45:51 UTC
Review Request 65806: Fixed a bug where quota headroom is
miscalculated.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/65806/
-----------------------------------------------------------
Review request for mesos, Benjamin Mahler and Michael Park.
Bugs: MESOS-8604
https://issues.apache.org/jira/browse/MESOS-8604
Repository: mesos
Description
-------
When calculating the quota headroom, we failed to consider
ancestor's reservation allocated to the child. This leads
to under-estimation of available headroom and excessive
resources being set aside for headroom. See MESOS-8604.
This patches fixes this issue by counting ancestor's reservation
allocated to the child as allocated-reservation even though
the child itself has no reservation.
Diffs
-----
src/master/allocator/mesos/hierarchical.cpp 80e0a347d6641c922cf7c62bef220ff5c8864597
Diff: https://reviews.apache.org/r/65806/diff/1/
Testing
-------
Thanks,
Meng Zhu
Re: Review Request 65806: Fixed a bug where quota headroom is
miscalculated.
Posted by Meng Zhu <mz...@mesosphere.io>.
> On Feb. 26, 2018, 6:41 p.m., Benjamin Mahler wrote:
> > Does this need to depend on the previous renaming patch? Makes cherry-picking a little messier
Nope. Removed the dependency.
- Meng
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/65806/#review198304
-----------------------------------------------------------
On Feb. 26, 2018, 6:46 p.m., Meng Zhu wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/65806/
> -----------------------------------------------------------
>
> (Updated Feb. 26, 2018, 6:46 p.m.)
>
>
> Review request for mesos, Benjamin Mahler and Michael Park.
>
>
> Bugs: MESOS-8604
> https://issues.apache.org/jira/browse/MESOS-8604
>
>
> Repository: mesos
>
>
> Description
> -------
>
> When calculating the quota headroom, we failed to consider
> ancestor's reservation allocated to the child. This leads
> to under-estimation of available headroom and excessive
> resources being set aside for headroom. See MESOS-8604.
>
> This patches fixes this issue by counting ancestor's reservation
> allocated to the child as allocated-reservation even though
> the child itself has no reservation.
>
>
> Diffs
> -----
>
> src/master/allocator/mesos/hierarchical.cpp 80e0a347d6641c922cf7c62bef220ff5c8864597
>
>
> Diff: https://reviews.apache.org/r/65806/diff/1/
>
>
> Testing
> -------
>
> make check
> Dedicated test in #65807
>
>
> Thanks,
>
> Meng Zhu
>
>
Re: Review Request 65806: Fixed a bug where quota headroom is
miscalculated.
Posted by Benjamin Mahler <bm...@apache.org>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/65806/#review198304
-----------------------------------------------------------
Does this need to depend on the previous renaming patch? Makes cherry-picking a little messier
- Benjamin Mahler
On Feb. 27, 2018, 12:45 a.m., Meng Zhu wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/65806/
> -----------------------------------------------------------
>
> (Updated Feb. 27, 2018, 12:45 a.m.)
>
>
> Review request for mesos, Benjamin Mahler and Michael Park.
>
>
> Bugs: MESOS-8604
> https://issues.apache.org/jira/browse/MESOS-8604
>
>
> Repository: mesos
>
>
> Description
> -------
>
> When calculating the quota headroom, we failed to consider
> ancestor's reservation allocated to the child. This leads
> to under-estimation of available headroom and excessive
> resources being set aside for headroom. See MESOS-8604.
>
> This patches fixes this issue by counting ancestor's reservation
> allocated to the child as allocated-reservation even though
> the child itself has no reservation.
>
>
> Diffs
> -----
>
> src/master/allocator/mesos/hierarchical.cpp 80e0a347d6641c922cf7c62bef220ff5c8864597
>
>
> Diff: https://reviews.apache.org/r/65806/diff/1/
>
>
> Testing
> -------
>
> make check
> Dedicated test in #65807
>
>
> Thanks,
>
> Meng Zhu
>
>
Re: Review Request 65806: Fixed a bug where quota headroom is
miscalculated.
Posted by Benjamin Mahler <bm...@apache.org>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/65806/#review198307
-----------------------------------------------------------
Ship it!
src/master/allocator/mesos/hierarchical.cpp
Lines 1677-1678 (original), 1677-1679 (patched)
<https://reviews.apache.org/r/65806/#comment278431>
Do we need the new variable? Looks like we just need to do the following in the equation below:
```
availableHeadroom -=
Resources::sum(reservationScalarQuantities) -
totalAllocatedReservationScalarQuantities;
```
src/master/allocator/mesos/hierarchical.cpp
Lines 1681-1688 (patched)
<https://reviews.apache.org/r/65806/#comment278432>
How about:
// Calculate total allocated reservations. Note that we need to ensure
// we count a reservation for "a" being allocated to "a/b", therefore
// we cannot simply loop over the reservations' roles.
src/master/allocator/mesos/hierarchical.cpp
Lines 1697-1698 (patched)
<https://reviews.apache.org/r/65806/#comment278433>
We could inline this to make it a little more succinct:
```
continue; // This role has no allocation.
```
- Benjamin Mahler
On Feb. 27, 2018, 2:46 a.m., Meng Zhu wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/65806/
> -----------------------------------------------------------
>
> (Updated Feb. 27, 2018, 2:46 a.m.)
>
>
> Review request for mesos, Benjamin Mahler and Michael Park.
>
>
> Bugs: MESOS-8604
> https://issues.apache.org/jira/browse/MESOS-8604
>
>
> Repository: mesos
>
>
> Description
> -------
>
> When calculating the quota headroom, we failed to consider
> ancestor's reservation allocated to the child. This leads
> to under-estimation of available headroom and excessive
> resources being set aside for headroom. See MESOS-8604.
>
> This patches fixes this issue by counting ancestor's reservation
> allocated to the child as allocated-reservation even though
> the child itself has no reservation.
>
>
> Diffs
> -----
>
> src/master/allocator/mesos/hierarchical.cpp 80e0a347d6641c922cf7c62bef220ff5c8864597
>
>
> Diff: https://reviews.apache.org/r/65806/diff/1/
>
>
> Testing
> -------
>
> make check
> Dedicated test in #65807
>
>
> Thanks,
>
> Meng Zhu
>
>
Re: Review Request 65806: Fixed a bug where quota headroom is
miscalculated.
Posted by Meng Zhu <mz...@mesosphere.io>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/65806/
-----------------------------------------------------------
(Updated Feb. 26, 2018, 6:46 p.m.)
Review request for mesos, Benjamin Mahler and Michael Park.
Changes
-------
Removed the unnecessary patch dependency.
Bugs: MESOS-8604
https://issues.apache.org/jira/browse/MESOS-8604
Repository: mesos
Description
-------
When calculating the quota headroom, we failed to consider
ancestor's reservation allocated to the child. This leads
to under-estimation of available headroom and excessive
resources being set aside for headroom. See MESOS-8604.
This patches fixes this issue by counting ancestor's reservation
allocated to the child as allocated-reservation even though
the child itself has no reservation.
Diffs
-----
src/master/allocator/mesos/hierarchical.cpp 80e0a347d6641c922cf7c62bef220ff5c8864597
Diff: https://reviews.apache.org/r/65806/diff/1/
Testing
-------
make check
Dedicated test in #65807
Thanks,
Meng Zhu