Default for backups in AtomicConfiguration

[Vladimir Ozerov]

Igniters,

I noted that in our AtomicConfiguration we have PARTITIONED cache with
backups=0. This appears to be very unreliable configuration which do not
survive node crashes.
For example, if user create AtomicLong with this config and the kill a
node, he could receive error message "Failed to find atomic with given
name" without explaining him that it might be lost.

Looks like we should either change defaults (e.g. backups=1 or
REPLICATED cache),
or improve our error messages.

Thoughts?

Vladimir.

[yzhdanov]

How about this one - Failed to find atomic with given name (increase
AtomicConfiguration.backups
count if it has been lost due to node leave)

?

--Yakov

2016-03-31 9:50 GMT+03:00 Vladimir Ozerov <[hidden email]>:

> Igniters,
>
> I noted that in our AtomicConfiguration we have PARTITIONED cache with
> backups=0. This appears to be very unreliable configuration which do not
> survive node crashes.
> For example, if user create AtomicLong with this config and the kill a
> node, he could receive error message "Failed to find atomic with given
> name" without explaining him that it might be lost.
>
> Looks like we should either change defaults (e.g. backups=1 or
> REPLICATED cache),
> or improve our error messages.
>
> Thoughts?
>
> Vladimir.
>

[dsetrakyan]

The message looks a little incomplete to me. How about:

“Failed to find an atomic structure with given name (increase
AtomicConfiguration.backups if it was lost due to a failed or left node):
*MyStructureName*”

D.

On Thu, Mar 31, 2016 at 2:56 AM, Yakov Zhdanov <[hidden email]> wrote:

> How about this one - Failed to find atomic with given name (increase
> AtomicConfiguration.backups
> count if it has been lost due to node leave)
>
> ?
>
> --Yakov
>
> 2016-03-31 9:50 GMT+03:00 Vladimir Ozerov <[hidden email]>:
>
> > Igniters,
> >
> > I noted that in our AtomicConfiguration we have PARTITIONED cache with
> > backups=0. This appears to be very unreliable configuration which do not
> > survive node crashes.
> > For example, if user create AtomicLong with this config and the kill a
> > node, he could receive error message "Failed to find atomic with given
> > name" without explaining him that it might be lost.
> >
> > Looks like we should either change defaults (e.g. backups=1 or
> > REPLICATED cache),
> > or improve our error messages.
> >
> > Thoughts?
> >
> > Vladimir.
> >
>

[yzhdanov]

Fine for me. Vladimir, can you please change it in master?

--Yakov

2016-03-31 19:52 GMT+03:00 Dmitriy Setrakyan <[hidden email]>:

> The message looks a little incomplete to me. How about:
>
> “Failed to find an atomic structure with given name (increase
> AtomicConfiguration.backups if it was lost due to a failed or left node):
> *MyStructureName*”
>
> D.
>
> On Thu, Mar 31, 2016 at 2:56 AM, Yakov Zhdanov <[hidden email]>
> wrote:
>
> > How about this one - Failed to find atomic with given name (increase
> > AtomicConfiguration.backups
> > count if it has been lost due to node leave)
> >
> > ?
> >
> > --Yakov
> >
> > 2016-03-31 9:50 GMT+03:00 Vladimir Ozerov <[hidden email]>:
> >
> > > Igniters,
> > >
> > > I noted that in our AtomicConfiguration we have PARTITIONED cache with
> > > backups=0. This appears to be very unreliable configuration which do
> not
> > > survive node crashes.
> > > For example, if user create AtomicLong with this config and the kill a
> > > node, he could receive error message "Failed to find atomic with given
> > > name" without explaining him that it might be lost.
> > >
> > > Looks like we should either change defaults (e.g. backups=1 or
> > > REPLICATED cache),
> > > or improve our error messages.
> > >
> > > Thoughts?
> > >
> > > Vladimir.
> > >
> >
>