"previous: Upgrading to 2.7.1":#v2_7_1
+ h3. WebDAV service uses @/var/cache@ for file content
+
+ @keep-web@ now stores copies of recently accessed data blocks in @/var/cache/arvados/keep@ instead of in memory. That directory will be created automatically. The default cache size is 10% of the filesystem size. Use the new @Collections.WebDAVCache.DiskCacheSize@ config to specify a different percentage or an absolute size.
+
+ If the previously supported @MaxBlockEntries@ config is present, remove it to avoid warning messages at startup.
+
+h3. Check MaxGatewayTunnels config
+
+If you use the LSF or Slurm dispatcher, ensure the new @API.MaxGatewayTunnels@ config entry is high enough to support the size of your cluster. See "LSF docs":{{site.baseurl}}/install/crunch2-lsf/install-dispatch.html#MaxGatewayTunnels or "Slurm docs":{{site.baseurl}}/install/crunch2-slurm/install-dispatch.html#MaxGatewayTunnels for details.
+
h2(#2_7_1). v2.7.1 (2023-12-12)
"previous: Upgrading to 2.7.0":#v2_7_0
# also effectively limited by MaxConcurrentRailsRequests (see
# below) because most controller requests proxy through to the
# RailsAPI service.
+ #
+ # HTTP proxies and load balancers downstream of arvados services
+ # should be configured to allow at least {MaxConcurrentRequest +
+ # MaxQueuedRequests + MaxGatewayTunnels} concurrent requests.
MaxConcurrentRequests: 64
# Maximum number of concurrent requests to process concurrently
# the incoming request queue before returning 503.
MaxQueueTimeForLockRequests: 2s
+ # Maximum number of active gateway tunnel connections. One slot
+ # is consumed by each "container shell" connection. If using an
+ # HPC dispatcher (LSF or Slurm), one slot is consumed by each
+ # running container. These do not count toward
+ # MaxConcurrentRequests.
+ MaxGatewayTunnels: 1000
+
# Fraction of MaxConcurrentRequests that can be "log create"
# messages at any given time. This is to prevent logging
# updates from crowding out more important requests.
# Time to cache manifests, permission checks, and sessions.
TTL: 300s
- # Block cache entries. Each block consumes up to 64 MiB RAM.
- MaxBlockEntries: 20
+ # Maximum amount of data cached in /var/cache/arvados/keep.
+ # Can be given as a percentage ("10%") or a number of bytes
+ # ("10 GiB")
+ DiskCacheSize: 10%
# Approximate memory limit (in bytes) for session cache.
#
# Note this applies to the in-memory representation of
# projects and collections -- metadata, block locators,
- # filenames, etc. -- excluding cached file content, which is
- # limited by MaxBlockEntries.
- MaxCollectionBytes: 100000000
+ # filenames, etc. -- not the file data itself (see
+ # DiskCacheSize).
+ MaxCollectionBytes: 100 MB
# Persistent sessions.
MaxSessions: 100
type WebDAVCacheConfig struct {
TTL Duration
- MaxBlockEntries int
- MaxCollectionBytes int64
+ DiskCacheSize ByteSizeOrPercent
+ MaxCollectionBytes ByteSize
MaxSessions int
}
MaxConcurrentRailsRequests int
MaxConcurrentRequests int
MaxQueuedRequests int
+ MaxGatewayTunnels int
MaxQueueTimeForLockRequests Duration
LogCreateRequestFraction float64
MaxKeepBlobBuffers int