disabling mercurial repositories

disabling mercurial repositories

[Roger Pau Monné]

Hello,

Last week we had a bit of an emergency when a web crawler started
indexing all our mercurial repositories on xenbits, as caused the load
on xenbits to go beyond what it can handle.

As a temporary solution we decided to remove access to mercurial
repositories, but the contents there are AFAIK only for historical
repositories, so we might consider completely removing access to
mercurial repositories.  This would however require migrating any
repository we care about to git.

I would like an opinion from committers as well as the broad community
whether shutting down mercurial repositories and migrating whatever we
care about is appropriate.  Otherwise we will need to implement some
throttling to mercurial accesses in order to avoid overloading
xenbits.

Thanks, Roger.

Re: disabling mercurial repositories

[Andrew Cooper]

On 21/06/2022 14:48, Roger Pau Monné wrote:
> Hello,
>
> Last week we had a bit of an emergency when a web crawler started
> indexing all our mercurial repositories on xenbits, as caused the load
> on xenbits to go beyond what it can handle.
>
> As a temporary solution we decided to remove access to mercurial
> repositories, but the contents there are AFAIK only for historical
> repositories, so we might consider completely removing access to
> mercurial repositories.  This would however require migrating any
> repository we care about to git.
>
> I would like an opinion from committers as well as the broad community
> whether shutting down mercurial repositories and migrating whatever we
> care about is appropriate.  Otherwise we will need to implement some
> throttling to mercurial accesses in order to avoid overloading
> xenbits.

IIRC, we'd mostly moved off hg onto git before moving to the Linux
Foundation, where git became mandatory.  Hg hasn't been the primary dev
tool for ages, and git has only got more ubiquitous in the meantime.

I'd suggest keeping hgweb disabled for now and see if anyone complains.

~Andrew

Re: disabling mercurial repositories

[George Dunlap]

[-- Attachment #1.1: Type: text/plain, Size: 1289 bytes --]



> On 21 Jun 2022, at 15:01, Andrew Cooper <Andrew.Cooper3@citrix.com> wrote:
> 
> On 21/06/2022 14:48, Roger Pau Monné wrote:
>> Hello,
>> 
>> Last week we had a bit of an emergency when a web crawler started
>> indexing all our mercurial repositories on xenbits, as caused the load
>> on xenbits to go beyond what it can handle.
>> 
>> As a temporary solution we decided to remove access to mercurial
>> repositories, but the contents there are AFAIK only for historical
>> repositories, so we might consider completely removing access to
>> mercurial repositories. This would however require migrating any
>> repository we care about to git.
>> 
>> I would like an opinion from committers as well as the broad community
>> whether shutting down mercurial repositories and migrating whatever we
>> care about is appropriate. Otherwise we will need to implement some
>> throttling to mercurial accesses in order to avoid overloading
>> xenbits.
> 
> IIRC, we'd mostly moved off hg onto git before moving to the Linux
> Foundation, where git became mandatory.  Hg hasn't been the primary dev
> tool for ages, and git has only got more ubiquitous in the meantime.
> 
> I'd suggest keeping hgweb disabled for now and see if anyone complains.

+ 1

 -George

[-- Attachment #1.2: Type: text/html, Size: 6718 bytes --]

[-- Attachment #2: Message signed with OpenPGP --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: disabling mercurial repositories

[Jan Beulich]

On 21.06.2022 15:48, Roger Pau Monné wrote:
> Last week we had a bit of an emergency when a web crawler started
> indexing all our mercurial repositories on xenbits, as caused the load
> on xenbits to go beyond what it can handle.
> 
> As a temporary solution we decided to remove access to mercurial
> repositories, but the contents there are AFAIK only for historical
> repositories, so we might consider completely removing access to
> mercurial repositories.  This would however require migrating any
> repository we care about to git.
> 
> I would like an opinion from committers as well as the broad community
> whether shutting down mercurial repositories and migrating whatever we
> care about is appropriate.  Otherwise we will need to implement some
> throttling to mercurial accesses in order to avoid overloading
> xenbits.

While I wouldn't strictly mind its shutting off or the disabling of
hgweb as was suggested in a reply, either would mean to me personally
that it wouldn't be easy enough anymore to warrant trying to hunt
down the origin of certain Linux side aspects in the 2.6.18-xen tree.
Admittedly me doing so has become increasingly rare over time ...

Jan

Re: disabling mercurial repositories

[Andrew Cooper]

On 22/06/2022 09:41, Jan Beulich wrote:
> On 21.06.2022 15:48, Roger Pau Monné wrote:
>> Last week we had a bit of an emergency when a web crawler started
>> indexing all our mercurial repositories on xenbits, as caused the load
>> on xenbits to go beyond what it can handle.
>>
>> As a temporary solution we decided to remove access to mercurial
>> repositories, but the contents there are AFAIK only for historical
>> repositories, so we might consider completely removing access to
>> mercurial repositories.  This would however require migrating any
>> repository we care about to git.
>>
>> I would like an opinion from committers as well as the broad community
>> whether shutting down mercurial repositories and migrating whatever we
>> care about is appropriate.  Otherwise we will need to implement some
>> throttling to mercurial accesses in order to avoid overloading
>> xenbits.
> While I wouldn't strictly mind its shutting off or the disabling of
> hgweb as was suggested in a reply, either would mean to me personally
> that it wouldn't be easy enough anymore to warrant trying to hunt
> down the origin of certain Linux side aspects in the 2.6.18-xen tree.
> Admittedly me doing so has become increasingly rare over time ...

We could convert that into a git repo (probably a branch on an existing
Linux.git to save most of the conversion work) and make it available via
gitweb if it's still useful?

~Andrew

Re: disabling mercurial repositories

[Jan Beulich]

On 22.06.2022 10:47, Andrew Cooper wrote:
> On 22/06/2022 09:41, Jan Beulich wrote:
>> On 21.06.2022 15:48, Roger Pau Monné wrote:
>>> Last week we had a bit of an emergency when a web crawler started
>>> indexing all our mercurial repositories on xenbits, as caused the load
>>> on xenbits to go beyond what it can handle.
>>>
>>> As a temporary solution we decided to remove access to mercurial
>>> repositories, but the contents there are AFAIK only for historical
>>> repositories, so we might consider completely removing access to
>>> mercurial repositories.  This would however require migrating any
>>> repository we care about to git.
>>>
>>> I would like an opinion from committers as well as the broad community
>>> whether shutting down mercurial repositories and migrating whatever we
>>> care about is appropriate.  Otherwise we will need to implement some
>>> throttling to mercurial accesses in order to avoid overloading
>>> xenbits.
>> While I wouldn't strictly mind its shutting off or the disabling of
>> hgweb as was suggested in a reply, either would mean to me personally
>> that it wouldn't be easy enough anymore to warrant trying to hunt
>> down the origin of certain Linux side aspects in the 2.6.18-xen tree.
>> Admittedly me doing so has become increasingly rare over time ...
> 
> We could convert that into a git repo (probably a branch on an existing
> Linux.git to save most of the conversion work) and make it available via
> gitweb if it's still useful?

If such a conversion would go cleanly enough, why not.

Jan