My database is taking an age to be available

[Paul Hinett]

Hi,

My data center shutdown my server without asking me first, once it
rebooted and raven started running again my database has taken nearly 2
hours so far loading and it's still not loaded.

I checked the resource monitor and can see it's writing to the RVN.log
file and a couple of tmp files, so it is doing something i just don't
know what.

Database has around 10 million docs.

I have other smaller databases which have loaded up just fine.

Paul

[Paul Hinett]

build 2104 by the way. i can upgrade if it will help, if it's safe to
force close raven that is.

[Paul Hinett]

Alot of my indexes are corrupted or no longer valid, i am resetting
them now.

[Mauro Servienti]

This is a huge problem, isn't it?

If a server, or a process, crashes for some reasons after the reboot/restart the biggest is the database the longest it takes to be available to users.

Why Lucene indexes get corrupted? On other products based on Lucene (such as Solr) this does not happens.

I'm asking because I do not see as an option the need to introduce replication in order to prevent a failure as the one Paul is facing.

This is particularly important when deploying to Azure where to lifecycle of the VM/Web/Worker/Role is totally out of control.

.m

[Oren Eini (Ayende Rahien)]

The problem here isn't that they are corrupted, they problem here is that we need to check them for corruption.

There also appears to be a separate issue regarding how we flush things to disk, which we are also investigating.