(DOCSP-2738): Updated MMAP caution and stable packages table. #3403

atsansone · 2018-09-08T01:26:48Z

@rkumar-mongo : This updates the warning at the bottom of this section and updates the table in this section. Links have been checked.

kay-kim

Please don't use the icon. Leave as |checkmark|

atsansone · 2018-09-10T17:28:39Z

@kay-kim : Fixed.

ravindk89

I think this needs another pass, especially around big/little endianness and the architecture support.

ravindk89 · 2018-09-11T18:46:00Z

source/includes/fact-mmapv1-big-endian.rst

-   MMAPv1 is not supported on big-endian architectures such as s390x.
-   MongoDB returns an error if you set MMAPv1 as the
-   storage engine on a big-endian system.
+   MMAPv1 is unsupported on big-endian and certain bi-endian


ppc64le is specifically little-endian, not bi-endian. s390x appears to be big endian as well. Structuring things this way makes it seem like s390x and PowerPC is bi-endian. Saying bi-endian implies we wouldn't support an OS built for Big Endian. As I understand it, the Power8 processors support either big or little endian, but not both at the same time.

Suggestion: "MMAPv1 is not supported on any big-endian architecture. MMAPv1 also does not support PowerPC." Unless we have a definitive list of other bi-endian or little-endian architectures we don't support, 'certain' casts too wide a net.

Made some revisions. Made it clear that ppc64le is bi-endian. Bi-endian does mean either/or and not both.

ravindk89 · 2018-09-11T18:46:29Z

source/includes/fact-mmapv1-big-endian.rst

+   MMAPv1 is unsupported on big-endian and certain bi-endian
+   architectures. This includes the IBM z/Architecture (``s390x``) and
+   PowerPC (``ppc64le``) platforms. MongoDB returns an error if you
+   set MMAPv1 as the storage engine on ``s390x`` and ``ppc64le``


What does MongoDB return for all the other big-endian architectures? This should still be generic.

Which other architectures can you install this on? None, to my understanding. I don't feel right getting into platforms we don't support just as CYA.

ravindk89 · 2018-09-11T18:53:32Z

source/includes/list-table-products-supported-architecture.rst


   * - Product
-     - x86_64/amd64
+     - x86_64


Why was amd64 removed? Did we drop support for it? I don't see that indicated in the source ticket. Our documentation still lists amd64 as a supported arch.

Because it is a synonym: https://en.wikipedia.org/wiki/X86-64

ravindk89 · 2018-09-11T18:55:16Z

source/includes/list-table-products-supported-architecture.rst

     - s390x
-     - POWER8 (little endian)


Same question - why was POWER8 removed? as far as I can tell, ppc64le was introduced with POWER8. It's feasible users will recognize one over the other.

One is the chipset, the other is the os compatibility. If we say POWER8, we should switch the others to Intel/AMD, zSeries and ARM. I'd leave this be. We use ppc64le in many other places, including the filenames of the binaries.

source/includes/fact-mmapv1-big-endian.rst

ravindk89 · 2018-09-11T18:59:40Z

source/includes/list-table-products-supported-architecture.rst

     - s390x
-     - POWER8 (little endian)
     - ARMv8-A



suggest + -> and later.

Agreed. Fixed.

ravindk89 · 2018-09-11T19:13:27Z

Note that the entire mmapv1 section is only valid for master. There is upcoming-4.2 wherein mmapv1 is toasted. Once we get through the cleanups, you'll need to 'forward-port' this ticket removing the mmapv1 references.

atsansone

@rkumar-mongo : Updated and back to you.

atsansone · 2018-09-14T14:29:14Z

source/includes/list-table-products-supported-architecture.rst


   * - Product
-     - x86_64/amd64
+     - x86_64


Because it is a synonym: https://en.wikipedia.org/wiki/X86-64

atsansone · 2018-09-14T15:04:24Z

source/includes/fact-mmapv1-big-endian.rst

+   MMAPv1 is unsupported on big-endian and certain bi-endian
+   architectures. This includes the IBM z/Architecture (``s390x``) and
+   PowerPC (``ppc64le``) platforms. MongoDB returns an error if you
+   set MMAPv1 as the storage engine on ``s390x`` and ``ppc64le``


Which other architectures can you install this on? None, to my understanding. I don't feel right getting into platforms we don't support just as CYA.

source/includes/fact-mmapv1-big-endian.rst

atsansone · 2018-09-14T19:13:58Z

source/includes/fact-mmapv1-big-endian.rst

-   MMAPv1 is not supported on big-endian architectures such as s390x.
-   MongoDB returns an error if you set MMAPv1 as the
-   storage engine on a big-endian system.
+   MMAPv1 is unsupported on big-endian and certain bi-endian


Made some revisions. Made it clear that ppc64le is bi-endian. Bi-endian does mean either/or and not both.

atsansone · 2018-09-14T20:22:53Z

source/includes/list-table-products-supported-architecture.rst

     - s390x
-     - POWER8 (little endian)


One is the chipset, the other is the os compatibility. If we say POWER8, we should switch the others to Intel/AMD, zSeries and ARM. I'd leave this be. We use ppc64le in many other places, including the filenames of the binaries.

atsansone · 2018-09-14T20:27:58Z

source/includes/list-table-products-supported-architecture.rst

     - s390x
-     - POWER8 (little endian)
     - ARMv8-A



Agreed. Fixed.

ravindk89

A few more things to tighten up. I'd also like someone from the Platforms (?) team to make sure we're clear on this.

ravindk89 · 2018-09-24T19:17:25Z

source/includes/fact-mmapv1-big-endian.rst

-   storage engine on a big-endian system.
+   MMAPv1 is unsupported on big-endian architectures like the IBM
+   z/Architecture (``s390x``) and certain bi-endian architectures like
+   the PowerPC (``ppc64le``) platform. MongoDB returns an error if you


We should keep this second section generic as well. Otherwise it implies MongoDB does not return errors for any other big-endian architecture, even if MMAPv1 isn't supported.

"MongoDB returns an error when starting a {{mongod}} using MMAPv1 on an unsupported architecture."

ravindk89 · 2018-09-24T19:19:25Z

source/includes/fact-mmapv1-big-endian.rst

+   z/Architecture (``s390x``) and certain bi-endian architectures like
+   the PowerPC (``ppc64le``) platform. MongoDB returns an error if you
+   set ``MMAPv1`` as the storage engine for either the ``s390x`` or
+   ``ppc64le`` platforms. MMAPv1 was deprecated in MongoDB 4.0.


I realised that this include is also brought in to /core/mmapv1. Having this note here is kind of redundant as that page has a big deprecation notice at the top. I think it's OK to remove it. There shouldn't be any surprises w.r.t. MMAPv1 at this point.

ravindk89 · 2018-09-24T19:20:17Z

source/includes/fact-mmapv1-big-endian.rst

-   MongoDB returns an error if you set MMAPv1 as the
-   storage engine on a big-endian system.
+   MMAPv1 is unsupported on big-endian architectures like the IBM
+   z/Architecture (``s390x``) and certain bi-endian architectures like


I think this ultimately begs the question of what other bi-endian or little-endian archs we throw errors on (vs archs we simply don't test against for official support). If it's only ppc64le that we throw errors for, we should just state that.

Fixed but forgot to push.

atsansone

Back to you, @rkumar-mongo

atsansone · 2018-09-25T16:05:18Z

source/includes/fact-mmapv1-big-endian.rst

-   MongoDB returns an error if you set MMAPv1 as the
-   storage engine on a big-endian system.
+   MMAPv1 is unsupported on big-endian architectures like the IBM
+   z/Architecture (``s390x``) and certain bi-endian architectures like


Fixed but forgot to push.

atsansone · 2018-09-25T16:08:37Z

source/includes/fact-mmapv1-big-endian.rst

+   z/Architecture (``s390x``) and certain bi-endian architectures like
+   the PowerPC (``ppc64le``) platform. MongoDB returns an error if you
+   set ``MMAPv1`` as the storage engine for either the ``s390x`` or
+   ``ppc64le`` platforms. MMAPv1 was deprecated in MongoDB 4.0.


ravindk89 · 2018-09-25T17:50:43Z

source/includes/list-table-products-supported-architecture.rst

     - ARMv8-A

-   * - MongoDB 3.4
+   * - MongoDB Community (3.4+)


[NP] this reverted to +. Change to "3.4 and later"

ravindk89

LGTM mod a few nits. Please pass to platforms to confirm we're clear here.

atsansone · 2018-09-29T06:01:15Z

@acmorrow : Finally got you access to this repo. Can you have a look at this change? The staged versions are noted in the first comment.

acmorrow

This LGTM for the part about MMAPv1 with a suggestion.

acmorrow · 2018-09-30T12:38:05Z

source/includes/fact-mmapv1-big-endian.rst

-   MMAPv1 is not supported on big-endian architectures such as s390x.
-   MongoDB returns an error if you set MMAPv1 as the
-   storage engine on a big-endian system.
+   :doc:`MMAPv1 </core/mmapv1>` is unsupported on all big-endian


Rather than trying to state where it isn't supported, why not just state that it is only supported on Intel x86 processors?

For instance, you don't state here that it isn't supported on ARM64. The actual logic that determines whether mmapv1 is here:

https://github.com/mongodb/mongo/blob/666545a2c01ad724912391ca44be36981d3ed0d2/SConstruct#L1829-L1836

acmorrow

MMAP changes LGTM.

ravindk89 · 2018-10-11T18:30:23Z

merged in by hand to avoid some merge conflicts: b01adfc

atsansone added the Copy Review label Sep 8, 2018

atsansone assigned ravindk89 Sep 8, 2018

atsansone requested a review from ravindk89 September 8, 2018 01:26

kay-kim suggested changes Sep 10, 2018

View reviewed changes

atsansone force-pushed the DOCSP-2738 branch from effeadd to c59f2d8 Compare September 10, 2018 17:28

ravindk89 suggested changes Sep 11, 2018

View reviewed changes

atsansone commented Sep 14, 2018

View reviewed changes

ravindk89 suggested changes Sep 24, 2018

View reviewed changes

Anthony Sansone added 2 commits September 25, 2018 11:11

(DOCSP-2738): Updated MMAP caution and stable packages table.

3f44c4f

Updated per @rkumar-mongo review mongodb#2

cddbf80

atsansone force-pushed the DOCSP-2738 branch from abd3af9 to cddbf80 Compare September 25, 2018 16:12

atsansone commented Sep 25, 2018

View reviewed changes

ravindk89 reviewed Sep 25, 2018

View reviewed changes

ravindk89 approved these changes Sep 25, 2018

View reviewed changes

atsansone requested a review from acmorrow September 29, 2018 06:00

atsansone assigned acmorrow and unassigned ravindk89 Sep 29, 2018

atsansone added Tech Review and removed Copy Review labels Sep 29, 2018

acmorrow suggested changes Sep 30, 2018

View reviewed changes

Update per @acmorrow feedback.

1de7346

atsansone force-pushed the DOCSP-2738 branch from 0a6ccc5 to 1de7346 Compare October 1, 2018 04:07

acmorrow approved these changes Oct 1, 2018

View reviewed changes

atsansone added RFM and removed Tech Review labels Oct 1, 2018

atsansone assigned ravindk89 and unassigned acmorrow Oct 1, 2018

ravindk89 closed this Oct 11, 2018

mongo-cr-bot pushed a commit that referenced this pull request Jun 22, 2023

DOCSP-30918 fix page (#3403)

1e10bd1

(DOCSP-2738): Updated MMAP caution and stable packages table. #3403

(DOCSP-2738): Updated MMAP caution and stable packages table. #3403

Uh oh!

Conversation

atsansone commented Sep 8, 2018

Uh oh!

kay-kim left a comment

Choose a reason for hiding this comment

Uh oh!

atsansone commented Sep 10, 2018

Uh oh!

ravindk89 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ravindk89 commented Sep 11, 2018

Uh oh!

atsansone left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ravindk89 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

atsansone left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ravindk89 left a comment

Choose a reason for hiding this comment

Uh oh!

atsansone commented Sep 29, 2018

Uh oh!

acmorrow left a comment

Choose a reason for hiding this comment