Which python API module?

[LenR]

I have used the lukecyca pyzabbix package for years, but just hit a possible bug, I happened to see on the Zabbix integrations page that their recommended package is the abubkov/py-zabbix. I wonder why that one is recommended, the lukecyca seems more popular.

[Markku]

Please tell more about the possible bug.

There is even discussion in GitHub issue that the packages are maybe merging: https://github.com/lukecyca/pyzabbix/issues/164

pyzabbix does not have a sender, maybe that's why Zabbix is "recommending" py-zabbix. (Yes, I hate that py-zabbix managed to select the same pyzabbix namespace...)

(And yes I'm using pyzabbix in my current projects)

Markku

[LenR]

I just opened an issue on the lukecyca project. I'm trying to httptest.create with a longer name than I'd ever had before. This name is 68 characters long and I got error

pyzabbix.api.ZabbixAPIException: ('Error -32602: Invalid params., Invalid parameter "/1/name": value is too long.', -32602)

I can create the web test via the Zabbix frontend, so it looks like a limit in the module, not Zabbix.

[Markku]

Ok, let's discuss that in the issue, please submit an example there.

Markku

[LenR]

z_url sanitized =

http://xxx-yyy-main.zzzz.zzzzzzzz.zzz/installation/default-en-US.htm

API call: minimally sanitized, no guarantee on the indention here, this worked with shorter url's.

web_resp = zapi.httptest.create(
{
    "name": z_url,
    "hostid": hostId,
    "delay": "240",
    "verify_host": "1",
    "verify_peer": "1",
    "steps": [
        {
            "name": "s1",
            "url": z_url,
            "status_codes": "200",
            "no": "1",
        }
    ],
    "tags": [
        {"tag": "Some Tag1", "value": "urlpath"},
        {"tag": "Application", "value": "Local Preciousness"},
    ],
}
)​

[Markku]

Have you tried

a) creating the "long" test in curl? That would mean that API is working but pyzabbix is not

b) creating the "long" test in GUI, reading it with pyzabbix, deleting in GUI, and recreating the test with the saved data in pyzabbix? (That would ensure that your input data is otherwise correct for the case)

Markku

[Markku]

For the case b), I think I have encountered some case where the error message from Zabbix was not actually correct, that's why I think it is useful to check the request data otherwise as well.

Markku

[Markku]

Btw what Zabbix version is this?

Markku

[LenR]

Zabbix is 6.0, I've not done the more difficult testing, it's been a LONG TIME since I tested with curl. I'll get to it eventually, but today is my Friday.

[Markku]

Fair enough. As a general comment, pyzabbix is not mangling the data much, and there are no practical string limits in Python, so my assumption based on the case description is that the problem is in Zabbix API, not pyzabbix. But let's see later.

Markku

[LenR]

I agree, I did try in py-zabbix and it gets the same issue.

I have support, I'll open a case :-)

[Markku]

My Zabbix 6.2.4 testhost truncates the web scenario name at 64 characters in GUI.

Markku

[LenR]

Oh, I hadn't noticed that, it worked but I didn't see the shorter name. I'm not sure if that is better or worse.

There is nothing to protect me form a use asking for 2 urls that have the same first 64 characters. Users are the root cause here...

[Markku]

But it's just the name, right? Not the URL

Markku

[LenR]

Right, but I was trying to use the url for the name, otherwise I'll have to invent a name.

[Markku]

Two hardest problems in IT:
1. Users
2. Naming things
3. Off-by-one errors



Markku