Not able to resume a stopped cloud on AWS instances

I have been working on installing solr cloud over AWS. As fresh run everything works fine and since i have used Hadoop as one of the dependencies it is understood that there is high availability.. So taking this base of though, i tried to stop the cloudera manager (essentially freezing hadoop , solr, other components). Then stop the instances come back next day to resume the working.. But this theory never works.. Below are the step by step things i do before i switch off and resume.

1. 7:55, created folders for every datanode and in ~/recovery directory and also checked each nodes health
2. copied the namenode current directory nn + dn from all 9 hosts with the help 2.*.sh script
3. Stopped the cloudera manager & ready to switch off the cluster
4. AT 8:04 cluster is stopped in cloudera manager. Made sure enough time is there between event 2 & 3 above.
5. Physically stopping the aws instances at 8:05.
6. Everything is stopped at 8:08
7. Starting all the nodes again.. at 8:12
8. started all all fine.. HDFS lost some blocks, some corrupted and some missing
9. Solr cloud failed.. totally because it is observed most of the blocks belong to solr cloud..

As you see i have taken all the precautions even i have redistributed the nn + dd which i have saved before switch off. But it did not work..

This i have failed the 4th time, and restoring the cloud is a painful procedure. Why i want to do this, i want to save the client some valuable money when we are not undergoing any testing.

I am still not sure why i can resume from my physical machine and not from aws.. And why only Solr looses bytes.