Bug 1636502

Summary: Unwanted repeated logs in self heal daemon logs
Product: [Community] GlusterFS Reporter: Shyamsundar <srangana>
Component: replicateAssignee: Ravishankar N <ravishankar>
Status: CLOSED CURRENTRELEASE QA Contact:
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: mainlineCC: atumball, bugs, ravishankar
Target Milestone: ---Keywords: Triaged
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: glusterfs-6.x Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of:
: 1636504 (view as bug list) Environment:
Last Closed: 2019-07-05 08:41:39 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Bug Depends On:    
Bug Blocks: 1636504    

Description Shyamsundar 2018-10-05 14:47:36 UTC 
Description of problem:

Run a test that needs self heal (say bring down a brick, touch 10 files in others and bring the brick back up online). The self heal logs are filled with the following log message,

[dict.c:2775:dict_get_str_boolean] (-->/usr/local/lib/glusterfs/6dev/xlator/protocol/client.so(+0x6d6c0) [0x7f3bc5ad76c0] -->/usr/local/lib/glusterfs/6dev/xlator/cluster/replicate.so(+0x489ff) [0x7f3bc581d9ff] -->/usr/local/lib/libglusterfs.so.0(dict_get_str_boolean+0xcf) [0x7f3bd3f0c81f] ) 0-dict: key buf-has-zeroes, integer type asked, has unsigned integer type [Invalid argument]

NOTE: Line numbers may be off depending on which commit in master I tested with

This is being generated as, POSIX is encoding a uint32 in the dict for buf-has-zeroes" and AFR is reading a boolean, from code inspection as below,

afr-self-heal-data.c -> __checksum_cbk -> dict_get_str_boolean(xdata, "buf-has-zeroes" ...

posix-inode-fd-ops.c -> posix_rchecksum -> dict_set_uint32(rsp_xdata, "buf-has-zeroes"

The correction needs type safety, and hence the bug.
Comment 1 Ravishankar N 2018-10-08 11:25:40 UTC 
https://review.gluster.org/#/c/glusterfs/+/21366/
Comment 2 Worker Ant 2018-10-08 11:26:02 UTC 
REVIEW: https://review.gluster.org/21366 (libglusterfs: change the validation type in dict_get_str_boolean) posted (#1) for review on master by Ravishankar N
Comment 3 Yaniv Kaul 2018-12-24 09:05:51 UTC 
The fix has caused others to pop up. For example:
[2018-12-23 23:52:37.333162] I [dict.c:2619:dict_get_str_boolean] (-->/build/install/lib/glusterfs/6dev/xlator/mgmt/glusterd.so(+0xb3626) [0x7fc032ef3626] -->/build/install/lib/glusterfs/6dev/xlator/mgmt/glusterd.so(+0xaf8db) [0x7fc032eef8db] -->/build/install/lib/libglusterfs.so.0(dict_get_str_boolean+0x170) [0x7fc03e282008] ) 0-dict: key nfs.disable, integer type asked, has string type [Invalid argument]

The reason being that nfs.disable (NFS_DISABLE_MAP_KEY) is set as string. For example:
if (dict_get_str_boolean(volinfo->dict, NFS_DISABLE_MAP_KEY, 1)) {
            ret = dict_set_dynstr_with_alloc(volinfo->dict, NFS_DISABLE_MAP_KEY,
                                             "on");


The correct thing of course is to introduce a sane boolean setter/getter and be done with all those hacks ...
Comment 4 Amar Tumballi 2019-07-05 08:41:39 UTC 
These warnings are not showing up no more in default logs.