B.39 store.*
Every field the catalogue defines under store, in path order. Each entry gives the field's definition and, under Carried by, every event type that writes it.
B.39.1 store.committed.cpus #
- Type:
uint[] - Asserted: no
- Carried in: the payload
- Defined in:
eventd.evman
The CPUs whose KMES rings eventd was reading when it shut down, one entry per
logical CPU in ascending order. Parallel to store.committed.sequences.
Carried by:
B.39.2 store.committed.sequences #
- Type:
uint[] - Asserted: no
- Carried in: the payload
- Defined in:
eventd.evman
For each CPU in store.committed.cpus, the highest event sequence of this
boot committed with no gap before it, taken during a graceful shutdown after
eventd had stopped reading the rings and committed everything it had read.
Parallel to store.committed.cpus. It is diagnostic only: when
eventd next starts it works out where to resume from the store itself, never
from this record.
0 means no unbroken run from sequence 1 exists for that CPU, as in
store.resume.sequences. When eventd cannot read its record of what was
committed, it leaves both arrays out of the shutdown record rather than
write zeroes (PEI-1394), so a 0 here is always a real value.
Carried by:
B.39.3 store.kind #
- Type:
str.enum - Values:
event·log·metric - Set: open
- Asserted: no
- Carried in: the payload
- Defined in:
eventd.evman
Which of eventd's stores a record is about: the sharded event store, the log
store or the metric store. On a storage error this is the store eventd found
corrupt and quarantined, and with store.shard it is the field worth alerting
on. The set is open because the database eventd keeps its own metadata in
may be reported on separately later; today eventd rebuilds that database
when it is damaged, and writes no record of it.
Carried by:
B.39.4 store.restarted #
- Type:
bool - Asserted: no
- Carried in: the payload
- Defined in:
eventd.evman
Whether this boot's events were already in the store when eventd started, which is to say eventd had already run during this boot and this start is a restart. It is false on eventd's first start in a boot. Counting the records where it is true answers how often eventd restarted during a boot, without inferring it from gaps.
A historical shard that cannot be opened at startup is left out of the decision, so on a damaged store a restart can be reported as false.
Carried by:
B.39.5 store.resume.cpus #
- Type:
uint[] - Asserted: no
- Carried in: the payload
- Defined in:
eventd.evman
The CPUs whose KMES rings eventd resumed reading at startup, one entry per
logical CPU in ascending order. Parallel to store.resume.sequences: entry
i of that array is where reading resumed for the CPU in entry i of this
one.
Carried by:
B.39.6 store.resume.sequences #
- Type:
uint[] - Asserted: no
- Carried in: the payload
- Defined in:
eventd.evman
For each CPU in store.resume.cpus, the highest event sequence of this boot
already committed with no gap before it, which is where eventd resumed
reading that CPU's ring. Parallel to store.resume.cpus.
0 is a value, not an absence. It means no unbroken run of committed events starting at sequence 1 exists for that CPU in this boot: nothing was committed yet, or the first events were lost. Events committed after a gap are not reflected here.
Carried by:
B.39.7 store.shard #
- Type:
uint - Asserted: no
- Carried in: the payload
- Defined in:
eventd.evman
The index of the event-store shard a record is about, counting from 0.
Present only when store.kind is event; the log and metric stores are not
sharded, so a record about either has no shard.
Carried by:
B.39.8 store.shard-count #
- Type:
uint - Asserted: no
- Carried in: the payload
- Defined in:
eventd.evman
How many active event-store shards eventd is running with, after resolving its configured shard count. Historical shards kept only for reading are not counted.
Carried by:
Generated from eventd.evman by pkm/tools/gen-events-book.py. Edit the fragment, not this page.