Downgrade from Database 7.1 to 7.0
Configuration changes
-
Replace the
flush-size
namespace configuration parameter with an explicitwrite-block-size 8M
. In Database 7.1 and later thewrite-block-size
was pinned to 8MiB. To reduce it, zeroize the drives on each downgraded node before restarting it. If you do not explicitly adjust thewrite-block-size
correctly, it will implicitly be assigned a 1MiB value, which is smaller than the 8MiB storage. This will manifest in the following error on startup:May 19 2024 17:26:14 GMT: INFO (drv_ssd): (drv_ssd.c:1056) /test/dev/xvdf has 5120 wblocks of size 1048576 May 19 2024 17:26:14 GMT: CRITICAL (drv_ssd): (drv_ssd.c:2390) /test/dev/xvdf: can't change write-block-size from 8388608 to 1048576 -
Replace your
post-write-cache
namespace configuration parameter with apost-write-queue
. If you keep yourwrite-block-size
at 8MiB, you can divide thepost-write-cache
by 8MiB to get the number of write-blocks to cache. -
Remove any configuration parameters added in Database 7.1 from your configuration file:
default-password-file
,indexes-memory-budget
,evict-indexes-memory-pct
,default-read-touch-ttl-pct
,auto-revive
,tls-refresh-period
,batch-max-requests
.