Open ayurchen opened 9 years ago
@byte Can we close this issue? Or should we document a way for manual cleanup, preferably without taking the node down?
@byte Can we close this issue? Or should we document a way for manual cleanup, preferably without taking the node down?
still incomplete
@byte I checked that the System Configuration tutorial has:
wsrep_provider_options="gcache.size=300M; gcache.page_size=300M"
If manual cleanup instructions are needed, I don't know how manual cleanup is done. Do you know?
@byte I checked that the System Configuration tutorial has:
wsrep_provider_options="gcache.size=300M; gcache.page_size=300M"
If manual cleanup instructions are needed, I don't know how manual cleanup is done. Do you know?
where is the current link as well
@byte Do you mean this link: https://galeracluster.com/library/training/tutorials/configuration.html
@byte Is there still something I must add to docs or can we close this issue?
Migrated from codership/galera#373
Customer is using Galera 3.25 with MySQL 5.6.23: galera-3-25.3.10-2.el6.x86_64 mysql-wsrep-client-5.6-5.6.23-25.10.el6.x86_64 mysql-wsrep-5.6-5.6.23-25.10.el6.x86_64 mysql-wsrep-server-5.6-5.6.23-25.10.el6.x86_64
Customer followed config example on page "http://galeracluster.com/documentation-webpages/configuration.html":
wsrep_provider_options="gcache.size=300M; gcache.page_size=1G"
"gcache" cleanup doesn't work, 148 gcache pages of 1 GB each have accumulated since Sep 2. Customer is approaching "disk full" and must take action.
In chat communication, I received this info from Codership: "The value of page_size parameter does not really matter, it just regulates how often file creation/deletion operation happens. but it makes little sense to have it bigger than the main cache size." (which is very plausible to me). I also was warned against removing gcache pages while the node is running.
Please: 1) Get the config example and that chat advice in sync, they contradict each other. 2) Check whether "size < page_size" may break the automatic cleanup. 3) Document a way for manual cleanup, preferably without taking the node down.