[checkmk-commits] Check_MK Git: check_mk: FIX zfsget: better filesystem selection and calculation of sizes

Module: check_mk Branch: master Commit: fbfba6eb0f6333e1f22d2908a4731c4f234ce1db URL: http://git.mathias-kettner.de/git/?p=check_mk.git;a=commit;h=fbfba6eb0f6333… Author: Goetz Golla <gg(a)mathias-kettner.de> Date: Fri Jun 13 16:00:05 2014 +0200 FIX zfsget: better filesystem selection and calculation of sizes zfsget now only inventorizes filesystems of type "filesystem". If a mountpoint is present several times in the zfsget section of the agent data (with different device name), and also in the df section, the entry in the zfsget section with the device name from the df section is taken for calculating the data. --- .werks/621 | 13 +++++++++++++ ChangeLog | 1 + checkman/zfsget | 6 +++--- checks/zfsget | 43 +++++++++++++++++++++++++++++++------------ 4 files changed, 48 insertions(+), 15 deletions(-) diff --git a/.werks/621 b/.werks/621 new file mode 100644 index 0000000..f0a8ead --- /dev/null +++ b/.werks/621 @@ -0,0 +1,13 @@ +Title: zfsget: better filesystem selection and calculation of sizes +Level: 2 +Component: checks +Version: 1.2.5i4 +Date: 1402667669 +Class: fix + +zfsget now only inventorizes filesystems of type "filesystem". + +If a mountpoint is present several times in the zfsget section of the agent +data (with different device name), and also in the df section, the entry +in the zfsget section with the device name from the df section is taken for +calculating the data. diff --git a/ChangeLog b/ChangeLog index b4b2e93..7c797c3 100644 --- a/ChangeLog +++ b/ChangeLog @@ -9,6 +9,7 @@ * 0619 adva_fsp_temp: new check to monitor temperature and temperature trends on ADVA scalable optical transport solutions * 0616 FIX: brocade.fan, brocade.power, brocade.temp: will now only discover services which are not marked as absent * 0992 FIX: zfs_arc_cache: returns OK even if values of arc meta are missing... + * 0621 FIX: zfsget: better filesystem selection and calculation of sizes... Multisite: * 0934 FIX: Logwatch messages with class unknown ( 'u' ) now displayed as WARN... diff --git a/checkman/zfsget b/checkman/zfsget index df2cd05..7b21ea3 100644 --- a/checkman/zfsget +++ b/checkman/zfsget @@ -17,7 +17,7 @@ description: filesystems live together in a common "pool". Each of the filesystems has its own amount of used space. But they share a common free space. If you want to compute the {percentage} of used/free space of a filesystem, - then you run into trouble. Let's take the following example: + you may run into trouble. Let's take the following example: Lets say we have a pool P of the size 10GB where the filesystems A and B live. Lets assume that A currently uses 2GB and B uses 7GB. That makes 1GB @@ -29,6 +29,6 @@ description: as 3GB with 2GB used (which makes 66.6%). That way all filesystems will become critical if the available pool space goes to zero. - Filesystems with a {quota} are handled differntly, however! Here the quota - is reported as the total size. If A would have a quota of 5GB then we report + Filesystems with a {quota} are handled differently, however. Here the quota + is reported as the total size. If A has a quota of 5GB, we report 2GB out of 5GB used. diff --git a/checks/zfsget b/checks/zfsget index 386e08f..c97e526 100644 --- a/checks/zfsget +++ b/checks/zfsget @@ -91,18 +91,40 @@ def parse_zfsget(info): if entry: entries.append(entry) + # parsed has the device name as key, because there may exist + # several device names per mount point, and we know only + # later which one to take parsed = {} for entry in entries: if entry["mountpoint"].startswith("/"): entry["is_pool"] = '/' not in entry["name"] - if entry['available'] != 0: - parsed[entry["mountpoint"]] = entry + if entry['available'] !=0 and entry['type'] == 'filesystem': + parsed[entry["name"]] = entry + # parsed_df and parsed_final have the mount point as key + parsed_final = {} if start_of_df != None: - zfsget_parse_df_info(parsed, info[start_of_df:]) - - return parsed - + parsed_df = zfsget_parse_df_info(parsed, info[start_of_df:]) + for mountpoint, entry_df in parsed_df.items(): + found = False + # for every mount point in the df section, if the device name + # is also present in the "parsed" variable, we take those data + for name, entry in parsed.items(): + if entry_df["name"] == name: + parsed_final[mountpoint] = entry + found = True + # if a mount point in the df section is not present in the + # parsed variable, we take the data from the df section + if not found: + parsed_final[mountpoint] = entry_df + else: + # if there is no df section we take the other data, + # should there be several device names per mount point, + # (not sure if this can happen) the last one wins + for name, entry in parsed.items(): + parsed_final[entry["mountpoint"]] = entry + + return parsed_final def zfsget_parse_df_info(entries, info): new_entries = {} @@ -112,11 +134,10 @@ def zfsget_parse_df_info(entries, info): device, kbytes, used, avail, percent, mountpoint = entry else: device, fs_type, kbytes, used, avail, percent, mountpoint = entry - # ignore entries already contained in zfsget and also - # entries for virtual filesystems (like swap) - if mountpoint.startswith("/") and mountpoint not in entries: + if mountpoint.startswith("/"): entry = {} total = int(kbytes) / 1024.0 + entry["name"] = device entry["total"] = total entry["used"] = int(used) / 1024.0 entry["available"] = total - entry["used"] @@ -138,9 +159,7 @@ def zfsget_parse_df_info(entries, info): drop.append(mountpoint) for mp in drop: del new_entries[mp] - entries.update(new_entries) - - + return new_entries def inventory_zfsget(info): mplist = []