Rocksolid Light

News from da outaworlds

mail  files  register  groups  login

Message-ID:  

BOFH excuse #401: Sales staff sold a product we don't offer.

comp / comp.mail.uucp / Re: Ensuring in-order execution

SubjectAuthor
* Ensuring in-order executionJohn Goerzen
`- Re: Ensuring in-order executionJohn Goerzen

1
Subject: Ensuring in-order execution
From: John Goerzen
Newsgroups: comp.mail.uucp
Organization: Alexandria NNCP news system
Date: Mon, 8 Nov 2021 14:53 UTC
Path: eternal-september.org!news.eternal-september.org!reader01.eternal-september.org!reader02.eternal-september.org!news.quux.org!alexnews.alexandria.complete.org!.POSTED!not-for-mail
From: jgoerzen@complete.org (John Goerzen)
Newsgroups: comp.mail.uucp
Subject: Ensuring in-order execution
Date: Mon, 8 Nov 2021 14:53:12 -0000 (UTC)
Organization: Alexandria NNCP news system
Message-ID: <slrnsoieeo.1n8qd.jgoerzen@slrnh.complete.org>
Injection-Date: Mon, 8 Nov 2021 14:53:12 -0000 (UTC)
Injection-Info: alexnews.alexandria.complete.org;
logging-data="494469"; mail-complaints-to="jgoerzen@complete.org"
User-Agent: slrn/1.0.3 (Linux)
View all headers

Hi,

I'm wondering if there are any generic programs out there to help ensure
in-order execution.

My particular use case: sending incremental backups over NNCP, though this is
the same problem as sending them over UUCP; namely, that nncp-exec/uuxqt aren't
able to make guarantees about executing commands in order, since they may arrive
via varying paths, etc.

I've worked around this in a couple of ways:

1) When using ZFS incrementals, zfs receive detects the out of order situation
and exits with an error.

2) When using tar or dar incrementals, I keep a "last processed" file on both
the sender and recipient, and part of the request references the sequence number
of the previous file, so the wrapper script can verify that before proceeding.

It's this situation #2 that I think must certainly have some generic solution,
as I keep seeming to reinvent it every few months. I'm sure that in the earlier
days of Unix and UUCP, this was a fairly common problem and it seems to me that
a generic solution must exist, but my searching has been futile so far.

Any ideas welcome.

Thanks,

John

Subject: Re: Ensuring in-order execution
From: John Goerzen
Newsgroups: comp.mail.uucp
Organization: Alexandria NNCP news system
Date: Tue, 14 Jun 2022 17:44 UTC
References: 1
Path: eternal-september.org!news.eternal-september.org!reader01.eternal-september.org!reader02.eternal-september.org!border1.nntp.dca1.giganews.com!nntp.giganews.com!weretis.net!feeder6.news.weretis.net!news.quux.org!alexnews.alexandria.complete.org!.POSTED!not-for-mail
From: jgoerzen@complete.org (John Goerzen)
Newsgroups: comp.mail.uucp
Subject: Re: Ensuring in-order execution
Date: Tue, 14 Jun 2022 17:44:00 -0000 (UTC)
Organization: Alexandria NNCP news system
Message-ID: <slrntahi70.27q8m.jgoerzen@slrnh.complete.org>
References: <slrnsoieeo.1n8qd.jgoerzen@slrnh.complete.org>
Injection-Date: Tue, 14 Jun 2022 17:44:00 -0000 (UTC)
Injection-Info: alexnews.alexandria.complete.org;
logging-data="346448"; mail-complaints-to="jgoerzen@complete.org"
User-Agent: slrn/1.0.3 (Linux)
Lines: 33
View all headers

I eventually decided to write something to solve this:

https://changelog.complete.org/archives/10368-fast-ordered-unixy-queues-over-nncp-and-syncthing-with-filespooler

On 2021-11-08, John Goerzen <jgoerzen@complete.org> wrote:
> Hi,
>
> I'm wondering if there are any generic programs out there to help ensure
> in-order execution.
>
> My particular use case: sending incremental backups over NNCP, though this is
> the same problem as sending them over UUCP; namely, that nncp-exec/uuxqt aren't
> able to make guarantees about executing commands in order, since they may arrive
> via varying paths, etc.
>
> I've worked around this in a couple of ways:
>
> 1) When using ZFS incrementals, zfs receive detects the out of order situation
> and exits with an error.
>
> 2) When using tar or dar incrementals, I keep a "last processed" file on both
> the sender and recipient, and part of the request references the sequence number
> of the previous file, so the wrapper script can verify that before proceeding.
>
> It's this situation #2 that I think must certainly have some generic solution,
> as I keep seeming to reinvent it every few months. I'm sure that in the earlier
> days of Unix and UUCP, this was a fairly common problem and it seems to me that
> a generic solution must exist, but my searching has been futile so far.
>
> Any ideas welcome.
>
> Thanks,
>
> John

1

rocksolid light 0.9.8
clearnet tor