DAOS-16477 pool: Rename Suspect state to Dead #15584
Conversation
Change the name to more closely reflect the underlying SWIM status, and reduce user confusion. An engine that has been marked DEAD by SWIM cannot participate in pool services, and has most likely already SIGKILL-ed itself. Features: pool Required-githooks: true Signed-off-by: Michael MacDonald <[email protected]>
|
Ticket title is 'Provide admin interface to query hanging engines after massive failure' |
github-actions
bot
added
the
priority
| @@ -162,8 +162,8 @@ enum daos_pool_info_bit { | |||
| DPI_ENGINES_ENABLED = 1ULL << 2, | |||
| /** true to include (in \a ranks) engines with some or all targets disabled (down). */ | |||
| DPI_ENGINES_DISABLED = 1ULL << 3, | |||
| /** true to include (in \a ranks) suspect engines. */ | |||
| DPI_ENGINES_SUSPECT = 1ULL << 4, | |||
| /** true to include (in \a ranks) engines marked DEAD by SWIM. */ | |||
There was a problem hiding this comment.
[Nit; not requesting changes] Maybe this degree of inaccuracy is acceptable, but I just want to clarify that these "dead" engines are actually the set of pool engines who are not "alive". In other words, they may also include engines absent from the SWIM membership.
There was a problem hiding this comment.
if that is the case then maybe suspect is a more accurate term?
There was a problem hiding this comment.
@liw: When would we have engines in the pool map that are absent from the SWIM membership, but aren't dead?
There was a problem hiding this comment.
When the MS has excluded the engines from the system (and the new system map has been distributed out), but the PS hasn't excluded the engines from the pool. In this state, from the pool's point of view, these engines are not strictly SWIM-DEAD, just unavailable---loosely speaking, "dead". (I think we should generally not differentiate between these two kinds of unavailability, SWIM-DEAD versus absent.)
Hmm, personally I don't think "suspect" is more accurate. :)
|
Test stage Functional on EL 8.8 completed with status FAILURE. https://build.hpdd.intel.com//job/daos-stack/job/daos/view/change-requests/job/PR-15584/1/execution/node/1210/log |
| @@ -162,8 +162,8 @@ enum daos_pool_info_bit { | |||
| DPI_ENGINES_ENABLED = 1ULL << 2, | |||
| /** true to include (in \a ranks) engines with some or all targets disabled (down). */ | |||
| DPI_ENGINES_DISABLED = 1ULL << 3, | |||
| /** true to include (in \a ranks) suspect engines. */ | |||
| DPI_ENGINES_SUSPECT = 1ULL << 4, | |||
| /** true to include (in \a ranks) engines marked DEAD by SWIM. */ | |||
There was a problem hiding this comment.
if that is the case then maybe suspect is a more accurate term?
|
Test failure appears to be https://daosio.atlassian.net/issues/DAOS-16829, which was introduced with the PR that added the suspect state. I could re-run without |
Features: pool Allow-unstable-test: true Required-githooks: true
|
I've merged master and re-pushed with the |
|
Test stage Functional on EL 8.8 completed with status FAILURE. https://build.hpdd.intel.com//job/daos-stack/job/daos/view/change-requests/job/PR-15584/2/execution/node/1210/log |
|
Test stage Functional Hardware Medium completed with status UNSTABLE. https://build.hpdd.intel.com/job/daos-stack/job/daos//view/change-requests/job/PR-15584/2/testReport/ |
|
Two known failures: |
Change the name to more closely reflect the underlying
SWIM status, and reduce user confusion. An engine that
has been marked DEAD by SWIM cannot participate in pool
services, and has most likely already SIGKILL-ed itself.
Features: pool
Required-githooks: true
Signed-off-by: Michael MacDonald [email protected]