incubator-s4-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Matthieu Morel (Commented) (JIRA)" <>
Subject [jira] [Commented] (S4-42) Optional recovery of expired and checkpointed PEs.
Date Tue, 31 Jan 2012 10:26:10 GMT


Matthieu Morel commented on S4-42:

I uploaded a solution in the S4-42 branch.

h4. Configuration:
_recoveryAfterExpiration_ can be specified for each prototype with a true|false value. Default
value is false: PEs will *not* be recovered after expiration.

h4. How it works:
When added to a cache, PEs keep the date of insertion, and this date is kept in the checkpoint.
Upon startup of an S4 node, if a checkpoint for the current PE instance if found, it is fetched
and deserialized. Then we evaluate recoveryAfterExpiration:
* true --> restore state from the checkpoint
* false --> compute expiration time from cache insertion + ttl. 
** If the PE is not expired, restore state. 
** Otherwise, do not restore previous state.

> Optional recovery of expired and checkpointed PEs.
> --------------------------------------------------
>                 Key: S4-42
>                 URL:
>             Project: Apache S4
>          Issue Type: New Feature
>    Affects Versions: 0.4
>            Reporter: Matthieu Morel
>             Fix For: 0.4
> S4 provides a caching mechanism for PEs, and PEs can be assigned a time to live. After
that time, the PE instance is removed from the cache and may be garbage collected.
> When combining with the checkpointing mechanism, expired but checkpointed PEs are restored
from a former checkpoint.
> This may or may not be desired, depending on the use case. Therefore we must provide
an option to restore - or not - the state of an expired but checkpointed PE.

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:!default.jspa
For more information on JIRA, see:


View raw message