After few months switch my Hadoop cluster from Cloudera CHD 4 to Hortowork data platform HDP 2.0, my cluster run smooth and stable but i think Hortonwork have a lot of thing to improve they product, hope that!
1. Install (with Linux enviroment)
• fail if prepair enviroment pharse is not correctly, no suggest for user to fix
• fail if ambari server and agent have time out, if your server have high load
• a lot of error don't have error code, hard to find and fix error
• some warning don't have any line of log out or log err
2. Service management
• slow service management
• unsupported chain service management in many case (ex: restart slave service if main service restart, after reconfig main service, ...)
• wrong service monitor info
• service down and up but history show in drashboard still OK, nothing about service down event show in history
• drashboad still show info of service ( status DOWN) that removed from cluster
• remove services in current cluster must manual stop, start A LOT OF service, it take my time
• Add host to cluster
• Block admin web gui and release after it done
• Remove host from cluster
• Must manual restart services in other host in cluster
• Move master component of service to other host in cluster
• Block admin web gui and release after it done
3. OS support: no automtic install for Ubuntu, a big mistake !
4. Document: poor, most of example only for RHEL, Centos
5. Community: small, after 2 year Hortonwork forum have < 2000 topics, a lot of them is HOW TO INSTALL !
1. Install (with Linux enviroment)
• fail if prepair enviroment pharse is not correctly, no suggest for user to fix
• fail if ambari server and agent have time out, if your server have high load
• a lot of error don't have error code, hard to find and fix error
• some warning don't have any line of log out or log err
2. Service management
• slow service management
• unsupported chain service management in many case (ex: restart slave service if main service restart, after reconfig main service, ...)
• wrong service monitor info
• service down and up but history show in drashboard still OK, nothing about service down event show in history
• drashboad still show info of service ( status DOWN) that removed from cluster
• remove services in current cluster must manual stop, start A LOT OF service, it take my time
• Add host to cluster
• Block admin web gui and release after it done
• Remove host from cluster
• Must manual restart services in other host in cluster
• Move master component of service to other host in cluster
• Block admin web gui and release after it done
3. OS support: no automtic install for Ubuntu, a big mistake !
4. Document: poor, most of example only for RHEL, Centos
5. Community: small, after 2 year Hortonwork forum have < 2000 topics, a lot of them is HOW TO INSTALL !
I wanted to thank you for this excellent read. we provide bouncy castle rental singapore at affordable charges. to know more visit our website.
ReplyDelete