avoid writing blocks to draining nodes ? #815
Labels
No labels
action
check-aws
action
discussion-needed
action
for-external-contributors
action
for-newcomers
action
more-info-needed
action
need-funding
action
triage-required
kind
correctness
kind
ideas
kind
improvement
kind
performance
kind
testing
kind
usability
kind
wrong-behavior
prio
critical
prio
low
scope
admin-api
scope
background-healing
scope
build
scope
documentation
scope
k8s
scope
layout
scope
metadata
scope
ops
scope
rpc
scope
s3-api
scope
security
scope
telemetry
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference: Deuxfleurs/garage#815
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Currently, when nodes are being removed from the layout, they continue receiving writes for metadata as per the consistency protocol. This means also that they will continue receiving the block_ref entries and the data blocks for the partitions they were responsible of. This is an issue as it can cause nodes to fill up quickly, which is especially problematic when we are removing nodes that are full to replace them with bigger nodes. We should probably find a way to avoid storing blocks on draining nodes, as long as we have a quorum of writes on other nodes.
Algorithm: try to have a quorum only on the current nodes. If we can not, use older nodes to achieve a quorum