what happens when caches ssd died

Discussion in 'General Questions' started by phamduyt, Jul 30, 2014.

  1. phamduyt

    phamduyt Bit Poster

    Messages:
    1
    We are using PCS and have 2SSDs for local caches and journal caches.
    I want to know what happens when caches ssd died?

     
  2. AnnaM

    AnnaM Kilo Poster

    Messages:
    13
    From what I know PCS should automatically detect failed drives and perform auto-recovery.
     
  3. akirov

    akirov Bit Poster

    Messages:
    6
    PStorage has two types of caching: CS write-journaling and Client read cache.

    If SSD for read caching is died, the client automatically stops using the cache and transparently turns off caching. After that, the client will continue working without cache until restart with cache options.

    If CS use SSD journaling, all write operations are going to SSD, than a background process commit the data form SSD to HDD. Losing CS journal means losing CS at all. Thus, if SSD, that stored some CS journals, is broken, the CSs become “failure”(broken). The outage of SSD leads to simultaneous outage of several corresponding CSs. It’s not a problem, because PStorage never store more than one replica per physical node.
     
    Last edited: Sep 2, 2014
  4. akirov

    akirov Bit Poster

    Messages:
    6
    Best Answer
    If any CS failed by some reason (HDD is broken, SSD is broken, HW is destroyed), PStorage detects it and automatically starts replication to recover.
     
  5. rickoneil

    rickoneil Bit Poster

    Messages:
    3
    I had a SSD journal fail, it marked the CS as "failed". I have my replication set at 4, so eventually it started moving chunks around to cover the missing drive. I just stopped the other CS on that node and rebuilt them as new, added them back to the cluster and then eventually the system started using the new CS drives.
     

Share This Page