You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@mesos.apache.org by Jason Zhou <ja...@gmail.com> on 2024/08/06 16:29:29 UTC
Review Request 75143: [device manager] persist state on device manager state change
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/75143/
-----------------------------------------------------------
Review request for mesos and Benjamin Mahler.
Repository: mesos
Description
-------
Currently we do not checkpoint the device access state of each cgroup
when the configure or reconfigure is called. Meaning that we have
no way of recovering a cgroup's device access state.
We will checkpoint state of the device manager whenever its state
is being changed to ensure that we can recover the most recent state
when necessary.
Diffs
-----
src/slave/containerizer/device_manager/device_manager.cpp e613323dc47a7980984426d37b6fc5cfc52dffe0
Diff: https://reviews.apache.org/r/75143/diff/1/
Testing
-------
In later patches of the stack, we were able to ensure that the state was persisted and was able to be recovered.
Thanks,
Jason Zhou
Re: Review Request 75143: [device manager] persist state on device manager state change
Posted by Benjamin Mahler <bm...@apache.org>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/75143/#review226777
-----------------------------------------------------------
src/slave/containerizer/device_manager/device_manager.cpp
Lines 234 (patched)
<https://reviews.apache.org/r/75143/#comment315074>
maybe just s/access_map/states/
src/slave/containerizer/device_manager/device_manager.cpp
Lines 237-239 (patched)
<https://reviews.apache.org/r/75143/#comment315073>
nit: 4 space indent on paren wrapping
src/slave/containerizer/device_manager/device_manager.cpp
Lines 238 (patched)
<https://reviews.apache.org/r/75143/#comment315076>
how about: s/cgroup_device_state/access/ ?
src/slave/containerizer/device_manager/device_manager.cpp
Lines 240 (patched)
<https://reviews.apache.org/r/75143/#comment315075>
maybe just s/message_instance/state/
src/slave/containerizer/device_manager/device_manager.cpp
Lines 250-251 (patched)
<https://reviews.apache.org/r/75143/#comment315077>
use std::move here, or manipulate directly:
```
CgroupDeviceAccessMessage state = &(*(access_map.mutable_device_access_per_cgroup()))[cgroup];
foreach (const Entry& entry, cgroup_device_state.allow_list) {
state->add_allow_list(stringify(entry));
}
foreach (const Entry& entry, cgroup_device_state.deny_list) {
state->add_deny_list(stringify(entry));
}
```
- Benjamin Mahler
On Aug. 6, 2024, 4:29 p.m., Jason Zhou wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/75143/
> -----------------------------------------------------------
>
> (Updated Aug. 6, 2024, 4:29 p.m.)
>
>
> Review request for mesos and Benjamin Mahler.
>
>
> Repository: mesos
>
>
> Description
> -------
>
> Currently we do not checkpoint the device access state of each cgroup
> when the configure or reconfigure is called. Meaning that we have
> no way of recovering a cgroup's device access state.
>
> We will checkpoint state of the device manager whenever its state
> is being changed to ensure that we can recover the most recent state
> when necessary.
>
>
> Diffs
> -----
>
> src/slave/containerizer/device_manager/device_manager.cpp e613323dc47a7980984426d37b6fc5cfc52dffe0
>
>
> Diff: https://reviews.apache.org/r/75143/diff/2/
>
>
> Testing
> -------
>
> In later patches of the stack, we were able to ensure that the state was persisted and was able to be recovered.
>
>
> Thanks,
>
> Jason Zhou
>
>
Re: Review Request 75143: [device manager] persist state on device manager state change
Posted by Benjamin Mahler <bm...@apache.org>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/75143/#review226778
-----------------------------------------------------------
src/slave/containerizer/device_manager/device_manager.cpp
Line 233 (original), 266 (patched)
<https://reviews.apache.org/r/75143/#comment315078>
probably s/persist/checkpoint/ would be more clear
- Benjamin Mahler
On Aug. 6, 2024, 4:29 p.m., Jason Zhou wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/75143/
> -----------------------------------------------------------
>
> (Updated Aug. 6, 2024, 4:29 p.m.)
>
>
> Review request for mesos and Benjamin Mahler.
>
>
> Repository: mesos
>
>
> Description
> -------
>
> Currently we do not checkpoint the device access state of each cgroup
> when the configure or reconfigure is called. Meaning that we have
> no way of recovering a cgroup's device access state.
>
> We will checkpoint state of the device manager whenever its state
> is being changed to ensure that we can recover the most recent state
> when necessary.
>
>
> Diffs
> -----
>
> src/slave/containerizer/device_manager/device_manager.cpp e613323dc47a7980984426d37b6fc5cfc52dffe0
>
>
> Diff: https://reviews.apache.org/r/75143/diff/2/
>
>
> Testing
> -------
>
> In later patches of the stack, we were able to ensure that the state was persisted and was able to be recovered.
>
>
> Thanks,
>
> Jason Zhou
>
>
Re: Review Request 75143: [device manager] checkpoint state on device manager state change
Posted by Benjamin Mahler <bm...@apache.org>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/75143/#review226785
-----------------------------------------------------------
Ship it!
Ship It!
- Benjamin Mahler
On Aug. 7, 2024, 4:55 p.m., Jason Zhou wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/75143/
> -----------------------------------------------------------
>
> (Updated Aug. 7, 2024, 4:55 p.m.)
>
>
> Review request for mesos and Benjamin Mahler.
>
>
> Repository: mesos
>
>
> Description
> -------
>
> Currently we do not checkpoint the device access state of each cgroup
> when the configure or reconfigure is called. Meaning that we have
> no way of recovering a cgroup's device access state.
>
> We will checkpoint state of the device manager whenever its state
> is being changed to ensure that we can recover the most recent state
> when necessary.
>
>
> Diffs
> -----
>
> src/slave/containerizer/device_manager/device_manager.cpp e613323dc47a7980984426d37b6fc5cfc52dffe0
>
>
> Diff: https://reviews.apache.org/r/75143/diff/5/
>
>
> Testing
> -------
>
> In later patches of the stack, we were able to ensure that the state was persisted and was able to be recovered.
>
>
> Thanks,
>
> Jason Zhou
>
>
Re: Review Request 75143: [device manager] checkpoint state on device manager state change
Posted by Jason Zhou <ja...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/75143/
-----------------------------------------------------------
(Updated Aug. 7, 2024, 4:55 p.m.)
Review request for mesos and Benjamin Mahler.
Repository: mesos
Description
-------
Currently we do not checkpoint the device access state of each cgroup
when the configure or reconfigure is called. Meaning that we have
no way of recovering a cgroup's device access state.
We will checkpoint state of the device manager whenever its state
is being changed to ensure that we can recover the most recent state
when necessary.
Diffs (updated)
-----
src/slave/containerizer/device_manager/device_manager.cpp e613323dc47a7980984426d37b6fc5cfc52dffe0
Diff: https://reviews.apache.org/r/75143/diff/5/
Changes: https://reviews.apache.org/r/75143/diff/4-5/
Testing
-------
In later patches of the stack, we were able to ensure that the state was persisted and was able to be recovered.
Thanks,
Jason Zhou
Re: Review Request 75143: [device manager] checkpoint state on device manager state change
Posted by Jason Zhou <ja...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/75143/
-----------------------------------------------------------
(Updated Aug. 7, 2024, 3:37 p.m.)
Review request for mesos and Benjamin Mahler.
Summary (updated)
-----------------
[device manager] checkpoint state on device manager state change
Repository: mesos
Description
-------
Currently we do not checkpoint the device access state of each cgroup
when the configure or reconfigure is called. Meaning that we have
no way of recovering a cgroup's device access state.
We will checkpoint state of the device manager whenever its state
is being changed to ensure that we can recover the most recent state
when necessary.
Diffs (updated)
-----
src/slave/containerizer/device_manager/device_manager.cpp e613323dc47a7980984426d37b6fc5cfc52dffe0
Diff: https://reviews.apache.org/r/75143/diff/3/
Changes: https://reviews.apache.org/r/75143/diff/2-3/
Testing
-------
In later patches of the stack, we were able to ensure that the state was persisted and was able to be recovered.
Thanks,
Jason Zhou