From ypiperea at cisco.com  Thu Jun  6 12:01:13 2019
From: ypiperea at cisco.com (Yohan Pipereau (ypiperea))
Date: Thu, 6 Jun 2019 12:01:13 +0000
Subject: [sysrepo-devel] Backup your running datastore
Message-ID: <DM6PR11MB34045F647568E1125CA00AC8C3170@DM6PR11MB3404.namprd11.prod.outlook.com>

Hi sysrepo community,

Andrej and I are part of the sweetcomb team which is leveraging sysrepo to expose a YANG interface for FD.io/VPP router.

First, many thanks to the sysrepo team for the work you are doing (and libyang, netopeer2 team if they are reading it too).

TLDR:
====

If sysrepo crashed, is there a way (even if it is unofficial) to get what was stored in RUNNING datastore before the crash ?

Longer explanations:
=================

Here is the global scenario:
So our goal is to put sysrepo in production systems with VPP router and to handle the case where VPP router and sysrepo would crash.
We already have a solution if VPP router crashes but we are lacking a solution in case sysrepo would crash.

I can read here http://www.sysrepo.org/static/doc/html/ds_page.html : "Be aware that changes in running datastore are not persistent and will be lost after the application or the system restarts - an explicit copy-config from running to startup datastore is needed to make the changes persistent after restarts".

We don't want to use STARTUP datastore. And the reason is simple:
-Before sysrepo crashes, RUNNING datastore had the same information as the router state but STARTUP datastore can hold a totally different state.
-After sysrepo crashes, we would like to have in RUNNING datastore the same state as the router, i.e. the state before the crash. We can't download this state from the router, because we would need to convert VPP state to YANG and it is not possible (it is only possible to convert YANG state to VPP state). So our solution would be to have the data which were in RUNNING datastore before the crash, and when sysrepo restart we could reapply this data. Do you know if such a thing exist ? I can see persist file in /etc/sysrepo/data, can it be used for this ?

Please feel free to ask for more details if I am not making myself clear,

Thanks in advance for your help, 

Yohan

From mvasko at cesnet.cz  Thu Jun 20 07:29:14 2019
From: mvasko at cesnet.cz (=?utf-8?q?Michal_Va=C5=A1ko?=)
Date: Thu, 20 Jun 2019 09:29:14 +0200
Subject: [sysrepo-devel] =?utf-8?q?Backup_your_running_datastore?=
In-Reply-To: <DM6PR11MB34045F647568E1125CA00AC8C3170@DM6PR11MB3404.namprd11.prod.outlook.com>
Message-ID: <275c-5d0b3600-2b-4562e500@127901821>

Hi Yohan,
yes, you can use the persist file for this (and it is unofficial). Just be aware that in some corner cases the file may be corrupted (sysrepod crashed after all).

Regards,
Michal

On Thursday, June 6, 2019 14:01 CEST, "Yohan Pipereau (ypiperea)" <ypiperea at cisco.com> wrote: 
 
> Hi sysrepo community,
> 
> Andrej and I are part of the sweetcomb team which is leveraging sysrepo to expose a YANG interface for FD.io/VPP router.
> 
> First, many thanks to the sysrepo team for the work you are doing (and libyang, netopeer2 team if they are reading it too).
> 
> TLDR:
> ====
> 
> If sysrepo crashed, is there a way (even if it is unofficial) to get what was stored in RUNNING datastore before the crash ?
> 
> Longer explanations:
> =================
> 
> Here is the global scenario:
> So our goal is to put sysrepo in production systems with VPP router and to handle the case where VPP router and sysrepo would crash.
> We already have a solution if VPP router crashes but we are lacking a solution in case sysrepo would crash.
> 
> I can read here http://www.sysrepo.org/static/doc/html/ds_page.html : "Be aware that changes in running datastore are not persistent and will be lost after the application or the system restarts - an explicit copy-config from running to startup datastore is needed to make the changes persistent after restarts".
> 
> We don't want to use STARTUP datastore. And the reason is simple:
> -Before sysrepo crashes, RUNNING datastore had the same information as the router state but STARTUP datastore can hold a totally different state.
> -After sysrepo crashes, we would like to have in RUNNING datastore the same state as the router, i.e. the state before the crash. We can't download this state from the router, because we would need to convert VPP state to YANG and it is not possible (it is only possible to convert YANG state to VPP state). So our solution would be to have the data which were in RUNNING datastore before the crash, and when sysrepo restart we could reapply this data. Do you know if such a thing exist ? I can see persist file in /etc/sysrepo/data, can it be used for this ?
> 
> Please feel free to ask for more details if I am not making myself clear,
> 
> Thanks in advance for your help, 
> 
> Yohan
 
 


From ypiperea at cisco.com  Thu Jun 20 08:53:14 2019
From: ypiperea at cisco.com (Yohan Pipereau (ypiperea))
Date: Thu, 20 Jun 2019 08:53:14 +0000
Subject: [sysrepo-devel] Backup your running datastore
In-Reply-To: <275c-5d0b3600-2b-4562e500@127901821>
References: <DM6PR11MB34045F647568E1125CA00AC8C3170@DM6PR11MB3404.namprd11.prod.outlook.com>
 <275c-5d0b3600-2b-4562e500@127901821>
Message-ID: <DM6PR11MB34049CD6F86C07664248F8B3C3E40@DM6PR11MB3404.namprd11.prod.outlook.com>

Hi Michal, 

Thanks for the information, this might be our best option for the moment.

Have a nice day,

Yohan

> -----Original Message-----
> From: Michal Va?ko <mvasko at cesnet.cz>
> Sent: jeudi 20 juin 2019 09:29
> To: Yohan Pipereau (ypiperea) <ypiperea at cisco.com>
> Cc: sysrepo-devel at sysrepo.org; Andrej Kozem??k
> <andrej.kozemcak at pantheon.tech>
> Subject: Re: Backup your running datastore
> 
> Hi Yohan,
> yes, you can use the persist file for this (and it is unofficial). Just be aware
> that in some corner cases the file may be corrupted (sysrepod crashed after
> all).
> 
> Regards,
> Michal
> 
> On Thursday, June 6, 2019 14:01 CEST, "Yohan Pipereau (ypiperea)"
> <ypiperea at cisco.com> wrote:
> 
> > Hi sysrepo community,
> >
> > Andrej and I are part of the sweetcomb team which is leveraging sysrepo to
> expose a YANG interface for FD.io/VPP router.
> >
> > First, many thanks to the sysrepo team for the work you are doing (and
> libyang, netopeer2 team if they are reading it too).
> >
> > TLDR:
> > ====
> >
> > If sysrepo crashed, is there a way (even if it is unofficial) to get what was
> stored in RUNNING datastore before the crash ?
> >
> > Longer explanations:
> > =================
> >
> > Here is the global scenario:
> > So our goal is to put sysrepo in production systems with VPP router and to
> handle the case where VPP router and sysrepo would crash.
> > We already have a solution if VPP router crashes but we are lacking a
> solution in case sysrepo would crash.
> >
> > I can read here http://www.sysrepo.org/static/doc/html/ds_page.html :
> "Be aware that changes in running datastore are not persistent and will be
> lost after the application or the system restarts - an explicit copy-config from
> running to startup datastore is needed to make the changes persistent after
> restarts".
> >
> > We don't want to use STARTUP datastore. And the reason is simple:
> > -Before sysrepo crashes, RUNNING datastore had the same information as
> the router state but STARTUP datastore can hold a totally different state.
> > -After sysrepo crashes, we would like to have in RUNNING datastore the
> same state as the router, i.e. the state before the crash. We can't download
> this state from the router, because we would need to convert VPP state to
> YANG and it is not possible (it is only possible to convert YANG state to VPP
> state). So our solution would be to have the data which were in RUNNING
> datastore before the crash, and when sysrepo restart we could reapply this
> data. Do you know if such a thing exist ? I can see persist file in
> /etc/sysrepo/data, can it be used for this ?
> >
> > Please feel free to ask for more details if I am not making myself clear,
> >
> > Thanks in advance for your help,
> >
> > Yohan
> 
>