diff options
author | Henning Baldersheim <balder@yahoo-inc.com> | 2018-04-06 13:03:43 +0200 |
---|---|---|
committer | Henning Baldersheim <balder@yahoo-inc.com> | 2018-04-06 13:17:47 +0200 |
commit | a52fc9fbee78e26e382feadf6f25a04a371eaab0 (patch) | |
tree | a2c258eb80cfd7c5fedf715477c02f3f585d48d8 /configdefinitions | |
parent | 5883ccf65ccc5ce6778704e3f4465a06eb77f391 (diff) |
Remove stor-devices
Diffstat (limited to 'configdefinitions')
-rw-r--r-- | configdefinitions/src/vespa/CMakeLists.txt | 2 | ||||
-rw-r--r-- | configdefinitions/src/vespa/stor-devices.def | 69 |
2 files changed, 0 insertions, 71 deletions
diff --git a/configdefinitions/src/vespa/CMakeLists.txt b/configdefinitions/src/vespa/CMakeLists.txt index f578d797f3a..d0a5132b90f 100644 --- a/configdefinitions/src/vespa/CMakeLists.txt +++ b/configdefinitions/src/vespa/CMakeLists.txt @@ -52,8 +52,6 @@ vespa_generate_config(configdefinitions slobroks.def) install_config_definition(slobroks.def cloud.config.slobroks.def) vespa_generate_config(configdefinitions specialtokens.def) install_config_definition(specialtokens.def vespa.configdefinition.specialtokens.def) -vespa_generate_config(configdefinitions stor-devices.def) -install_config_definition(stor-devices.def vespa.config.storage.stor-devices.def) vespa_generate_config(configdefinitions stor-distribution.def) install_config_definition(stor-distribution.def vespa.config.content.stor-distribution.def) vespa_generate_config(configdefinitions stor-filestor.def) diff --git a/configdefinitions/src/vespa/stor-devices.def b/configdefinitions/src/vespa/stor-devices.def deleted file mode 100644 index bab67cc76af..00000000000 --- a/configdefinitions/src/vespa/stor-devices.def +++ /dev/null @@ -1,69 +0,0 @@ -# Copyright 2017 Yahoo Holdings. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root. -namespace=vespa.config.storage - -## Root directory for all files related to this storage node. -## Will typically be "$VESPA_HOME/var/db/vespa/vds/<cluster>/<nodetype>/<index>" -## Within this directory the "disks/d<dirindex>" directories/symlinks will -## contain the actual data stored by storage. -root_folder string restart - -## List of disks to store the storage data in. Make an empty directory for -## each disk_path[index] if it does not exist yet, and make a symlink from each -## "<root_folder>/disks/d<index>" to disk_path[index] if it doesn't exist yet. -disk_path[] string restart - -## Capacity of the disks on this node. How much data and load will each disk -## get relative to the other disks on this node. The default disk capacity is -## 1.0. If disk capacity array size does not match disk counts, it will use the -## values that exist, and use 1.0 for the rest. -disk_capacity[] double restart - -## The device mapper is used to detect what storage directories map to what -## partitions and disks, such that then a disk event arise, it arise for all -## directories using that disk. Storage has two different device mappers. One -## that fakes the devices, assuming all VDS directories use unique partitions -## and disks, and a real one trying to detect what is actually being used. -## -## Advantage of simple one is that one can fake multiple disks, which may be -## useful in testing. Problem being that then storage gives clues that it has -## added X bytes to a partition, other directories on the same partition will -## not be adjusted. Thus, one might want to cache stat results less due to -## numbers being incorrect. -## -## If using a real mapper, disk events such as IO failures will mark all -## directories using partitions on that disk bad, rather than just the directory -## containing the file which we had trouble using. We dont detect real physical -## disks. Currently we detect whether they use the same *nix device or not. -device_mapper enum { SIMPLE_MAPPER, REAL_MAPPER } default=REAL_MAPPER restart - -## Whether or not to fail directories on same partition and/or disk as another -## directory on failures. Our goal is to return directory errors on failures -## only affecting one directory, partition failures for errors only affecting a -## single partition and so on. In reality, however, it is very hard to know if -## the problem is partition or disk related for instance. Both a corrupt -## filesystem and a bad disk may show the same IO error to storage. We -## recommend having this enabled, so we can attempt to do logic, but it is -## possible to turn it of, in case you see the logic doesnt work for you (and -## that you get more directories than needed marked down and that troubles -## your system) -fail_multiple_directories_on_common_events bool default=true restart - -## Storage stats file systems regularily to check how full they are. To avoid -## constantly calling this system call storage may cache the result and validate -## changes once in a while. The always policy always calls the real statfs -## function when asked for disk fullness. This may be useful in system testing -## where you want to test disk full situations (but not this policy). The stat -## once policy does a single stat upon start, and depends on storage to give -## clues about changes in disk usage to report correct results (currently not -## recommended since we dont guarantuee the clues are 100% correct). The period -## policy asks for each Nth request, and so does the dynamic policy, but the -## latter one has a dynamic period, being large when disk has low fill rate, but -## small with high fillrate. The dynamic policy is recommended as it doesnt -## require exact change clues requires few stat calls as long as the disks -## arent close to full. -statfs_policy enum { STAT_ALWAYS, STAT_ONCE, STAT_PERIOD, STAT_DYNAMIC } default=STAT_DYNAMIC restart - -## If set to non-null, overrides the code default period for stat period and -## dynamic policies. See PartitionMonitor class for details. -statfs_period int default=0 restart - |