Hi I have been doing some testing to see how a database(netscapedb) will catch up with replicated changes when the server has been shutdown and/or broken.
My test is very basic: Shutdown master2 Add an entry to netscapedb on master1 Bring up master2 Tail error log for replication messages and check console on master2 for presence of data.
I know this is imprecise but my observation has been so far that if master2 was down for a short period of time it will automatically catch up but if it were down for a longer period of time (more than an hour) and I bring it up, new updates are not automatically send. I have to click on send updates in the 389-console for the latest changes to be reflected.
Is there more definitive rules that govern when a database will be updated automatically and when the process becomes manual?
Best Regards
________________________________________________________________________ In order to protect our email recipients, Betfair Group use SkyScan from MessageLabs to scan all Incoming and Outgoing mail for viruses.
________________________________________________________________________
Gerrard Geldenhuis wrote:
Hi
I have been doing some testing to see how a database(netscapedb) will catch up with replicated changes when the server has been shutdown and/or broken.
My test is very basic:
Shutdown master2
Add an entry to netscapedb on master1
Bring up master2
Tail error log for replication messages and check console on master2 for presence of data.
I know this is imprecise but my observation has been so far that if master2 was down for a short period of time it will automatically catch up but if it were down for a longer period of time (more than an hour) and I bring it up, new updates are not automatically send. I have to click on send updates in the 389-console for the latest changes to be reflected.
How long do you wait? Replication should wait at most 5 minutes.
Is there more definitive rules that govern when a database will be updated automatically and when the process becomes manual?
Replication uses an exponential backoff strategy if the consumer is down. That is, it will wait 1 second, try again, then wait 2 seconds, try again, then wait 4 seconds, try again, etc. until it hits 5 minutes.
Best Regards
In order to protect our email recipients, Betfair Group use SkyScan from MessageLabs to scan all Incoming and Outgoing mail for viruses.
-- 389 users mailing list 389-users@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/389-users
Replication uses an exponential backoff strategy if the consumer is down. That is, it will wait 1 second, try again, then wait 2 seconds, try again, then wait 4 seconds, try again, etc. until it hits 5 minutes.
hmmm, I probably did not wait long enough...
I have enabled replication monitoring on all my servers at the moment.
Is there any log that specifically log the exponential back-off and the current backup-off waiting time? Alternatively can that value be read somewhere? If not I believe it would be useful to have in the logs and I could raise a feature request in bugzilla.
Does the exponential backoff reset after 5 minutes back to 1 second or does it stay at 5 minute interval checking?
Regards
________________________________________________________________________ In order to protect our email recipients, Betfair Group use SkyScan from MessageLabs to scan all Incoming and Outgoing mail for viruses.
________________________________________________________________________
Gerrard Geldenhuis wrote:
Replication uses an exponential backoff strategy if the consumer is down. That is, it will wait 1 second, try again, then wait 2 seconds, try again, then wait 4 seconds, try again, etc. until it hits 5 minutes.
hmmm, I probably did not wait long enough...
I have enabled replication monitoring on all my servers at the moment.
Is there any log that specifically log the exponential back-off and the current backup-off waiting time?
Not a log file. You can turn on the replication log level, but that will spew out a lot of unrelated messages too.
Alternatively can that value be read somewhere? If not I believe it would be useful to have in the logs and I could raise a feature request in bugzilla.
There are many attributes used for monitoring replication that you can read over LDAP http://www.redhat.com/docs/manuals/dir-server/8.2/cli/html/Core_Server_Confi...
Why do you need to know how long it will be? You could monitor nsDS5ReplicaLastInitEnd, nsDS5ReplicaLastInitStart, nsDS5ReplicaLastInitStatus, nsDS5ReplicaLastUpdateEnd, nsDS5ReplicaLastUpdateStart, nsDS5ReplicaLastUpdateStatus, nsDS5ReplicaUpdateInProgress to see if replication is occuring. If the consumer is down, the InitStatus or UpdateStatus will have some error like 81 or 91 indicating a connection could not be made. If you read that status, and you know the consumer is up, this means the server is waiting, and if you don't want to wait, you can just initiate an update. You do this by setting nsDS5ReplicaUpdateSchedule to a value that is not the current time e.g. nsDS5ReplicaUpdateSchedule: 0000-0001 0 then immediately set the value back to "replicate all the time" e.g. nsDS5ReplicaUpdateSchedule: 0000-2359 0123456 That will "wake up" the replication thread and make it send over all updates immediately.
Does the exponential backoff reset after 5 minutes back to 1 second or does it stay at 5 minute interval checking?
After a successful connection, it will reset back to 1 second.
Regards
In order to protect our email recipients, Betfair Group use SkyScan from MessageLabs to scan all Incoming and Outgoing mail for viruses.
-- 389 users mailing list 389-users@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/389-users
389-users@lists.fedoraproject.org
-
Gerrard Geldenhuis -
Rich Megginson