Wang Shilong
Wang Shilong
Add fault injection to pool upgrading, container upgrading will be aborted if fault injection is inserted, use this to verify pool upgrading could be resumed upon pool restarting Signed-off-by: Wang...
Now we have two different fault domains: rank and node. daos_obj_generate_oid_by_rf() use default fault domain which might be wrong. Eg, creating container(with rf_lvl:1) Introduce another helper daos_obj_generate_oid_by_rf2() which is aware...
1. switch one global lock to per bucket lock. to reduce lock contention for multiple threads. 2. switch spinlock to read/write lock on the client side, add following adjustment to...
daos_eq_lib_init() might set FI_OFI_RXM_USE_SRX, re-check if server is set or not. Otherwise daos_init()/fini() called several times behaviour might be confusing. Required-githooks: true Signed-off-by: Wang Shilong
If system domain_nr don't require group size, return error directly. And if system's targets could not meet group size * groups number, we should return error rather than search for...
Whenever we want to access pool and container pointer in daos client stack, looking a per-proccess handle hash table by daos handle is required. Performance could not scale well for...
The cluster has an UP state target, and the object's layout may have an extending target. Calling the pl_obj_layout_contains interface in the reclaim process essentially means that the variable i...
1. move init thread from main xstream to sys xstream 2. first detected meta SSD will be selected as sys device to store SMD, vos db pool blobid is stored...
Required-githooks: true Signed-off-by: Wang Shilong
If container is going to be dead, set @sc_stopping firstly, and later lookup will fail, then wait existed container services exit. ### Before requesting gatekeeper: * [ ] Two review...