Data migration testing for 0.6.1

We had a spirited time with the data migration during update
last time 
osaf.us was updated.  What things do we want to do
differently for the 
osaf.us 0.6.0.1 to 0.6.1 update?

I haven't seen "data migration" testing on 0.6.1
test plans (sorry if I 
missed them), so I'm sorta assuming that osaf.us-focused
testing will 
again be the driver for testing of the Cosmo data migration
features. 
That's perfectly fine; that's one of the reasons osaf.us is
there and 
I'm happy to be heavily involved.

When would people like to see a first migrated-data test
instance of 
osaf.us be made available?

How is QA thinking of the migration testing?  Who's on
point?  Should we 
document our migration test plan, and when?

I think we are likely to have the same approach to data
migration 
testing: part automated, part manual.  Automated testing
will use 
Mikeal's ICS-comparison script, and manual testing will
revolve around 
volunteers looking at their private data in a migrated-data
test 
instance during dedicated IRC sessions.

-- Jared
_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

Hi Jared,
Migration testing is definitely on our radar.  In fact we
just  
discussed about it and having a migration test plan in
today's QA  
staff meeting.
We haven't forgotten the pains from our last experience.

See further comments line..

On Apr 10, 2007, at 2:46 PM, Jared Rhine wrote:

> We had a spirited time with the data migration during
update last  
> time osaf.us was updated.  What things do we want to do
differently  
> for the osaf.us 0.6.0.1 to 0.6.1 update?

Have a migration test plan for one.
>
> I haven't seen "data migration" testing on
0.6.1 test plans (sorry  
> if I missed them), so I'm sorta assuming that
osaf.us-focused  
> testing will again be the driver for testing of the
Cosmo data  
> migration features. That's perfectly fine; that's one
of the  
> reasons osaf.us is there and I'm happy to be heavily
involved.
>
Yes, osaf.us-focused testing will indeed be the driver.

> When would people like to see a first migrated-data
test instance  
> of osaf.us be made available?
Ted just confirmed that Cosmo 0.6.1 is feature complete
(barring any  
compatibility bug fixes). Dev unit testing is underway this
week. So  
lets target on getting the first migrated test instance of
osaf.us to  
be available next week, Monday or Tuesday. We could then use
that for  
Wednesday's IRC QA session.
In the meantime we will run the migration scripts on the
limited data  
we have on qacosmo to filter out any obvious bugs.


>
> How is QA thinking of the migration testing?  Who's on
point?   
> Should we document our migration test plan, and when?

We probably need to sit down with you one of these days and
discuss  
about what things we would like to be covered, above and
beyond what  
is already on the 0.6.1 test plan. Mikeal is your point of
contact  
for migration.
>
> I think we are likely to have the same approach to data
migration  
> testing: part automated, part manual.  Automated
testing will use  
> Mikeal's ICS-comparison script, and manual testing will
revolve  
> around volunteers looking at their private data in a
migrated-data  
> test instance during dedicated IRC sessions.
Yes, we are definitely planning a few IRC test sessions
around 0.6.1  
and we should cover migration testing during that.


> -- Jared
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

Hi Jared,
Migration testing is definitely on our radar.  In fact we
just  
discussed about it and having a migration test plan in
today's QA  
staff meeting.
We haven't forgotten the pains from our last experience.

See further comments line..

On Apr 10, 2007, at 2:46 PM, Jared Rhine wrote:

> We had a spirited time with the data migration during
update last  
> time osaf.us was updated.  What things do we want to do
differently  
> for the osaf.us 0.6.0.1 to 0.6.1 update?

Have a migration test plan for one.
>
> I haven't seen "data migration" testing on
0.6.1 test plans (sorry  
> if I missed them), so I'm sorta assuming that
osaf.us-focused  
> testing will again be the driver for testing of the
Cosmo data  
> migration features. That's perfectly fine; that's one
of the  
> reasons osaf.us is there and I'm happy to be heavily
involved.
>
Yes, osaf.us-focused testing will indeed be the driver.

> When would people like to see a first migrated-data
test instance  
> of osaf.us be made available?
Ted just confirmed that Cosmo 0.6.1 is feature complete
(barring any  
compatibility bug fixes). Dev unit testing is underway this
week. So  
lets target on getting the first migrated test instance of
osaf.us to  
be available next week, Monday or Tuesday. We could then use
that for  
Wednesday's IRC QA session.
In the meantime we will run the migration scripts on the
limited data  
we have on qacosmo to filter out any obvious bugs.


>
> How is QA thinking of the migration testing?  Who's on
point?   
> Should we document our migration test plan, and when?

We probably need to sit down with you one of these days and
discuss  
about what things we would like to be covered, above and
beyond what  
is already on the 0.6.1 test plan. Mikeal is your point of
contact  
for migration.
>
> I think we are likely to have the same approach to data
migration  
> testing: part automated, part manual.  Automated
testing will use  
> Mikeal's ICS-comparison script, and manual testing will
revolve  
> around volunteers looking at their private data in a
migrated-data  
> test instance during dedicated IRC sessions.
Yes, we are definitely planning a few IRC test sessions
around 0.6.1  
and we should cover migration testing during that.


> -- Jared
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

> Yes, we are definitely planning a few IRC test sessions
around  
> 0.6.1 and we should cover migration testing during
that.

For the first session next week I have this suggestion.

Rather than migrating all the osaf.us data we have jared
host a clean  
cosmo 0.6.0 and for the first 20 minutes of the session we
load it up  
with our accounts and data from Chandler and other clients.
Then we have jared bring down the server, run his migration
script  
and update to 0.6.1.
Then we use the rest of the time in the session to find any
issues  
that may have happened during the migration.

With this process we find all the bugs that pertain
specifically to  
the same dataset migrated with the same subscription URLs
rather than  
testing all our data that's been migrated to different URLs
on the  
migrated instance. This would have flushed out a lot of the
issues we  
didn't catch last time in the migration testing process
until we went  
to production.

-Mikeal 

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

With the transition from the old dual-fork ICS+XML shares to
the new  
EIMML-based shares, it's going to have to be Chandler
clients that  
migrate the data.  Cosmo doesn't understand the old XML
format that  
Chandler's been using to store all non-event-related data. 
I guess  
there are really two aspects to this migration:  one which
copies the  
old ICS+XML files into an 0.6.1 Cosmo instance, and then one
where  
the person responsible for a given collection uses their
Chandler to  
convert ICS+XML to EIMML (and then unpublishes the original
dual-fork  
collection).

On Apr 10, 2007, at 3:42 PM, Mikeal Rogers wrote:

>
>> Yes, we are definitely planning a few IRC test
sessions around  
>> 0.6.1 and we should cover migration testing during
that.
>
> For the first session next week I have this
suggestion.
>
> Rather than migrating all the osaf.us data we have
jared host a  
> clean cosmo 0.6.0 and for the first 20 minutes of the
session we  
> load it up with our accounts and data from Chandler and
other clients.
> Then we have jared bring down the server, run his
migration script  
> and update to 0.6.1.
> Then we use the rest of the time in the session to find
any issues  
> that may have happened during the migration.
>
> With this process we find all the bugs that pertain
specifically to  
> the same dataset migrated with the same subscription
URLs rather  
> than testing all our data that's been migrated to
different URLs on  
> the migrated instance. This would have flushed out a
lot of the  
> issues we didn't catch last time in the migration
testing process  
> until we went to production.
>
> -Mikeal
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

> Yes, we are definitely planning a few IRC test sessions
around  
> 0.6.1 and we should cover migration testing during
that.

For the first session next week I have this suggestion.

Rather than migrating all the osaf.us data we have jared
host a clean  
cosmo 0.6.0 and for the first 20 minutes of the session we
load it up  
with our accounts and data from Chandler and other clients.
Then we have jared bring down the server, run his migration
script  
and update to 0.6.1.
Then we use the rest of the time in the session to find any
issues  
that may have happened during the migration.

With this process we find all the bugs that pertain
specifically to  
the same dataset migrated with the same subscription URLs
rather than  
testing all our data that's been migrated to different URLs
on the  
migrated instance. This would have flushed out a lot of the
issues we  
didn't catch last time in the migration testing process
until we went  
to production.

-Mikeal 

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

With the transition from the old dual-fork ICS+XML shares to
the new  
EIMML-based shares, it's going to have to be Chandler
clients that  
migrate the data.  Cosmo doesn't understand the old XML
format that  
Chandler's been using to store all non-event-related data. 
I guess  
there are really two aspects to this migration:  one which
copies the  
old ICS+XML files into an 0.6.1 Cosmo instance, and then one
where  
the person responsible for a given collection uses their
Chandler to  
convert ICS+XML to EIMML (and then unpublishes the original
dual-fork  
collection).

On Apr 10, 2007, at 3:42 PM, Mikeal Rogers wrote:

>
>> Yes, we are definitely planning a few IRC test
sessions around  
>> 0.6.1 and we should cover migration testing during
that.
>
> For the first session next week I have this
suggestion.
>
> Rather than migrating all the osaf.us data we have
jared host a  
> clean cosmo 0.6.0 and for the first 20 minutes of the
session we  
> load it up with our accounts and data from Chandler and
other clients.
> Then we have jared bring down the server, run his
migration script  
> and update to 0.6.1.
> Then we use the rest of the time in the session to find
any issues  
> that may have happened during the migration.
>
> With this process we find all the bugs that pertain
specifically to  
> the same dataset migrated with the same subscription
URLs rather  
> than testing all our data that's been migrated to
different URLs on  
> the migrated instance. This would have flushed out a
lot of the  
> issues we didn't catch last time in the migration
testing process  
> until we went to production.
>
> -Mikeal
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

You bring up a good point Morgen. There are more subtleties
to test  
in this migration than the previous.

Jared, Mikeal, Adam, Morgen, Ted: can we have a meeting
tomorrow or  
Thursday(after the cosmo meeting at 3 pm) to go over all the
cases we  
would like to cover during migration?
We can develop the migration test plan based on that.


On Apr 10, 2007, at 3:56 PM, Morgen Sagen wrote:

> With the transition from the old dual-fork ICS+XML
shares to the  
> new EIMML-based shares, it's going to have to be
Chandler clients  
> that migrate the data.  Cosmo doesn't understand the
old XML format  
> that Chandler's been using to store all
non-event-related data.  I  
> guess there are really two aspects to this migration: 
one which  
> copies the old ICS+XML files into an 0.6.1 Cosmo
instance, and then  
> one where the person responsible for a given collection
uses their  
> Chandler to convert ICS+XML to EIMML (and then
unpublishes the  
> original dual-fork collection).
>
> On Apr 10, 2007, at 3:42 PM, Mikeal Rogers wrote:
>
>>
>>> Yes, we are definitely planning a few IRC test
sessions around  
>>> 0.6.1 and we should cover migration testing
during that.
>>
>> For the first session next week I have this
suggestion.
>>
>> Rather than migrating all the osaf.us data we have
jared host a  
>> clean cosmo 0.6.0 and for the first 20 minutes of
the session we  
>> load it up with our accounts and data from Chandler
and other  
>> clients.
>> Then we have jared bring down the server, run his
migration script  
>> and update to 0.6.1.
>> Then we use the rest of the time in the session to
find any issues  
>> that may have happened during the migration.
>>
>> With this process we find all the bugs that pertain
specifically  
>> to the same dataset migrated with the same
subscription URLs  
>> rather than testing all our data that's been
migrated to different  
>> URLs on the migrated instance. This would have
flushed out a lot  
>> of the issues we didn't catch last time in the
migration testing  
>> process until we went to production.
>>
>> -Mikeal
>> _______________________________________________
>> cosmo-dev mailing list
>> cosmo-devlists.osafoundation.org
>> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev
>
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

You bring up a good point Morgen. There are more subtleties
to test  
in this migration than the previous.

Jared, Mikeal, Adam, Morgen, Ted: can we have a meeting
tomorrow or  
Thursday(after the cosmo meeting at 3 pm) to go over all the
cases we  
would like to cover during migration?
We can develop the migration test plan based on that.


On Apr 10, 2007, at 3:56 PM, Morgen Sagen wrote:

> With the transition from the old dual-fork ICS+XML
shares to the  
> new EIMML-based shares, it's going to have to be
Chandler clients  
> that migrate the data.  Cosmo doesn't understand the
old XML format  
> that Chandler's been using to store all
non-event-related data.  I  
> guess there are really two aspects to this migration: 
one which  
> copies the old ICS+XML files into an 0.6.1 Cosmo
instance, and then  
> one where the person responsible for a given collection
uses their  
> Chandler to convert ICS+XML to EIMML (and then
unpublishes the  
> original dual-fork collection).
>
> On Apr 10, 2007, at 3:42 PM, Mikeal Rogers wrote:
>
>>
>>> Yes, we are definitely planning a few IRC test
sessions around  
>>> 0.6.1 and we should cover migration testing
during that.
>>
>> For the first session next week I have this
suggestion.
>>
>> Rather than migrating all the osaf.us data we have
jared host a  
>> clean cosmo 0.6.0 and for the first 20 minutes of
the session we  
>> load it up with our accounts and data from Chandler
and other  
>> clients.
>> Then we have jared bring down the server, run his
migration script  
>> and update to 0.6.1.
>> Then we use the rest of the time in the session to
find any issues  
>> that may have happened during the migration.
>>
>> With this process we find all the bugs that pertain
specifically  
>> to the same dataset migrated with the same
subscription URLs  
>> rather than testing all our data that's been
migrated to different  
>> URLs on the migrated instance. This would have
flushed out a lot  
>> of the issues we didn't catch last time in the
migration testing  
>> process until we went to production.
>>
>> -Mikeal
>> _______________________________________________
>> cosmo-dev mailing list
>> cosmo-devlists.osafoundation.org
>> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev
>
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

The expectation here is that most people will be using old,
already  
released, Chandler clients through the next migration and
that they  
will still be using the old sharing methods.

Moving people between sharing accounts in Chandler hasn't
been  
discussed, and won't be fully addressed for the 0.6.1
release because  
most of the dogfooders won't be using EIMML capable
Chandlers against  
osaf.us, and cosmo isn't responsible for migrating that data
anyway.

We will need to have a plan for instructing our bleeding
edge users,  
I'm lookin' at you Andre : ), on how to migrate their shares
to the  
new sharing format, and for the Preview release of Chandler
we'll  
need a plan for instructing everyone on how to do the same.

EIMML testing is perhaps the largest part of them cosmo
0.6.1 release  
testing, but migration with EIMML is not part of that.

-Mikeal

On Apr 10, 2007, at 3:56 PM, Morgen Sagen wrote:

> With the transition from the old dual-fork ICS+XML
shares to the  
> new EIMML-based shares, it's going to have to be
Chandler clients  
> that migrate the data.  Cosmo doesn't understand the
old XML format  
> that Chandler's been using to store all
non-event-related data.  I  
> guess there are really two aspects to this migration: 
one which  
> copies the old ICS+XML files into an 0.6.1 Cosmo
instance, and then  
> one where the person responsible for a given collection
uses their  
> Chandler to convert ICS+XML to EIMML (and then
unpublishes the  
> original dual-fork collection).
>
> On Apr 10, 2007, at 3:42 PM, Mikeal Rogers wrote:
>
>>
>>> Yes, we are definitely planning a few IRC test
sessions around  
>>> 0.6.1 and we should cover migration testing
during that.
>>
>> For the first session next week I have this
suggestion.
>>
>> Rather than migrating all the osaf.us data we have
jared host a  
>> clean cosmo 0.6.0 and for the first 20 minutes of
the session we  
>> load it up with our accounts and data from Chandler
and other  
>> clients.
>> Then we have jared bring down the server, run his
migration script  
>> and update to 0.6.1.
>> Then we use the rest of the time in the session to
find any issues  
>> that may have happened during the migration.
>>
>> With this process we find all the bugs that pertain
specifically  
>> to the same dataset migrated with the same
subscription URLs  
>> rather than testing all our data that's been
migrated to different  
>> URLs on the migrated instance. This would have
flushed out a lot  
>> of the issues we didn't catch last time in the
migration testing  
>> process until we went to production.
>>
>> -Mikeal
>> _______________________________________________
>> cosmo-dev mailing list
>> cosmo-devlists.osafoundation.org
>> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev
>
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev


_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

The expectation here is that most people will be using old,
already  
released, Chandler clients through the next migration and
that they  
will still be using the old sharing methods.

Moving people between sharing accounts in Chandler hasn't
been  
discussed, and won't be fully addressed for the 0.6.1
release because  
most of the dogfooders won't be using EIMML capable
Chandlers against  
osaf.us, and cosmo isn't responsible for migrating that data
anyway.

We will need to have a plan for instructing our bleeding
edge users,  
I'm lookin' at you Andre : ), on how to migrate their shares
to the  
new sharing format, and for the Preview release of Chandler
we'll  
need a plan for instructing everyone on how to do the same.

EIMML testing is perhaps the largest part of them cosmo
0.6.1 release  
testing, but migration with EIMML is not part of that.

-Mikeal

On Apr 10, 2007, at 3:56 PM, Morgen Sagen wrote:

> With the transition from the old dual-fork ICS+XML
shares to the  
> new EIMML-based shares, it's going to have to be
Chandler clients  
> that migrate the data.  Cosmo doesn't understand the
old XML format  
> that Chandler's been using to store all
non-event-related data.  I  
> guess there are really two aspects to this migration: 
one which  
> copies the old ICS+XML files into an 0.6.1 Cosmo
instance, and then  
> one where the person responsible for a given collection
uses their  
> Chandler to convert ICS+XML to EIMML (and then
unpublishes the  
> original dual-fork collection).
>
> On Apr 10, 2007, at 3:42 PM, Mikeal Rogers wrote:
>
>>
>>> Yes, we are definitely planning a few IRC test
sessions around  
>>> 0.6.1 and we should cover migration testing
during that.
>>
>> For the first session next week I have this
suggestion.
>>
>> Rather than migrating all the osaf.us data we have
jared host a  
>> clean cosmo 0.6.0 and for the first 20 minutes of
the session we  
>> load it up with our accounts and data from Chandler
and other  
>> clients.
>> Then we have jared bring down the server, run his
migration script  
>> and update to 0.6.1.
>> Then we use the rest of the time in the session to
find any issues  
>> that may have happened during the migration.
>>
>> With this process we find all the bugs that pertain
specifically  
>> to the same dataset migrated with the same
subscription URLs  
>> rather than testing all our data that's been
migrated to different  
>> URLs on the migrated instance. This would have
flushed out a lot  
>> of the issues we didn't catch last time in the
migration testing  
>> process until we went to production.
>>
>> -Mikeal
>> _______________________________________________
>> cosmo-dev mailing list
>> cosmo-devlists.osafoundation.org
>> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev
>
> _______________________________________________
> cosmo-dev mailing list
> cosmo-devlists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev


_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

On Apr 10, 2007, at 4:06 PM, Aparna Kadakia wrote:

> Jared, Mikeal, Adam, Morgen, Ted: can we have a meeting
tomorrow or  
> Thursday(after the cosmo meeting at 3 pm) to go over
all the cases  
> we would like to cover during migration?
> We can develop the migration test plan based on that.

I am booked solid tomorrow but am open Thursday except for
the Cosmo  
engineering meeting at 2pm/bug council at 3pm

Ted
_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev

On Apr 10, 2007, at 4:06 PM, Aparna Kadakia wrote:

> Jared, Mikeal, Adam, Morgen, Ted: can we have a meeting
tomorrow or  
> Thursday(after the cosmo meeting at 3 pm) to go over
all the cases  
> we would like to cover during migration?
> We can develop the migration test plan based on that.

I am booked solid tomorrow but am open Thursday except for
the Cosmo  
engineering meeting at 2pm/bug council at 3pm

Ted
_______________________________________________
cosmo-dev mailing list
cosmo-devlists.osafoundation.org
http://lists.osafoundation.org/mailman/listinfo/cosmo-d
ev