[Glass] startStone -N

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

[Glass] startStone -N

Johan Brichau-3
Hi there,

When I need to start a stone using the -N parameter (because of corrupt tranlogs after inadvertent shutdown), do I need to restart the stone afterwards?
In other words: does the -N parameter have any influence beyond the startup of the stone?

Furthermore, when the stone has a startup error on a tranlog because of a bad EOF. Are there (easy) methods to try to repair the tranlog?
Even though we loose just the transactions after the last checkpoint.

thx
Johan
_______________________________________________
Glass mailing list
[hidden email]
http://lists.gemtalksystems.com/mailman/listinfo/glass
Reply | Threaded
Open this post in threaded view
|

Re: [Glass] startStone -N

James Foster-9
Hi Johan,

On Jun 2, 2014, at 11:32 AM, Johan Brichau <[hidden email]> wrote:

> Hi there,
>
> When I need to start a stone using the -N parameter (because of corrupt tranlogs after inadvertent shutdown), do I need to restart the stone afterwards?
> In other words: does the -N parameter have any influence beyond the startup of the stone?

Actually, you should not need to use -N and and inadvertent shutdown should not corrupt tranlogs.

But, no, the -N command line argument to the startstone command does not have any impact after the startup completes. The only difference is that the usual process of checking for a clean shutdown and attempting to reply transaction logs is skipped. Once the stone is running it is in normal mode.

The -R (start in restore mode) does leave the repository in a different state after the startup, and this can be used in conjunction with -N but that is optional.

> Furthermore, when the stone has a startup error on a tranlog because of a bad EOF. Are there (easy) methods to try to repair the tranlog?

In general, an incomplete transaction log record should be ignored during tranlog replay and the stone should read and apply every complete record up to end (or incomplete record). No repair should be needed.

> Even though we loose just the transactions after the last checkpoint.

The only lost transaction should be the incomplete one, and the system should not have returned from the #’commit’ or #’commitTransaction’ method until the write completed so the application should not have given any indication that the commit completed. This is why you should print receipts or send email AFTER the transaction finishes.

This is somewhat more detailed that the response I gave to Mariano a couple weeks ago and I’ll add some comments to that thread as well.

James

>
> thx
> Johan
> _______________________________________________
> Glass mailing list
> [hidden email]
> http://lists.gemtalksystems.com/mailman/listinfo/glass

_______________________________________________
Glass mailing list
[hidden email]
http://lists.gemtalksystems.com/mailman/listinfo/glass